Showing 51–100 of 123 results for author: Fox, E

The Atacama Cosmology Telescope: Delensed Power Spectra and Parameters

Authors: Dongwon Han, Neelima Sehgal, Amanda MacInnis, Alexander van Engelen, Blake D. Sherwin, Mathew S. Madhavacheril, Simone Aiola, Nicholas Battaglia, James A. Beall, Daniel T. Becker, Erminia Calabrese, Steve K. Choi, Omar Darwish, Edward V. Denison, Mark J. Devlin, Jo Dunkley, Simone Ferraro, Anna E. Fox, Matthew Hasselfield, J. Colin Hill, Gene C. Hilton, Matt Hilton, Renée Hložek, Johannes Hubmayr, John P. Hughes , et al. (17 additional authors not shown)

Abstract: We present LCDM cosmological parameter constraints obtained from delensed microwave background power spectra. Lensing maps from a subset of DR4 data from the Atacama Cosmology Telescope (ACT) are used to undo the lensing effect in ACT spectra observed at 150 and 98 GHz. At 150 GHz, we remove the lensing distortion with an effective efficiency of 30% (TT), 30% (EE), 26% (TE) and 20% (BB); this resu… ▽ More We present LCDM cosmological parameter constraints obtained from delensed microwave background power spectra. Lensing maps from a subset of DR4 data from the Atacama Cosmology Telescope (ACT) are used to undo the lensing effect in ACT spectra observed at 150 and 98 GHz. At 150 GHz, we remove the lensing distortion with an effective efficiency of 30% (TT), 30% (EE), 26% (TE) and 20% (BB); this results in detections of the delensing effect at 8.7 sigma (TT), 5.1 sigma (EE), 2.6 sigma (TE), and 2.4 sigma (BB) significance. The combination of 150 and 98 GHz TT, EE, and TE delensed spectra is well fit by a standard LCDM model. We also measure the shift in best-fit parameters when fitting delensed versus lensed spectra; while this shift does not inform our ability to measure cosmological parameters, it does provide a three-way consistency check among the lensing inferred from the best-fit parameters, the lensing in the CMB power spectrum, and the reconstructed lensing map. This shift is predicted to be zero when fitting with the correct model since both lensed and delensed spectra originate from the same region of sky. Fitting with a LCDM model and marginalizing over foregrounds, we find that the shift in cosmological parameters is consistent with zero. Our results show that gravitational lensing of the microwave background is internally consistent within the framework of the standard cosmological model. △ Less

Submitted 13 November, 2020; v1 submitted 28 July, 2020; originally announced July 2020.

Comments: 29 pages, 17 figures, version matches that accepted by JCAP

Journal ref: JCAP, Issue 01, article id. 031 (2021)

The Atacama Cosmology Telescope: DR5 maps of 18,000 square degrees of the microwave sky from ACT 2008-2018 data

Authors: Sigurd Naess, Simone Aiola, Jason E. Austermann, Nick Battaglia, James A. Beall, Daniel T. Becker, Richard J. Bond, Erminia Calabrese, Steve K. Choi, Nicholas F. Cothard, Kevin T. Crowley, Omar Darwish, Rahul Datta, Edward V. Denison, Mark Devlin, Cody J. Duell, Shannon M. Duff, Adriaan J. Duivenvoorden, Jo Dunkley, Rolando Dünner, Anna E. Fox, Patricio A. Gallardo, Mark Halpern, Dongwon Han, Matthew Hasselfield , et al. (37 additional authors not shown)

Abstract: This paper presents a maximum-likelihood algorithm for combining sky maps with disparate sky coverage, angular resolution and spatially varying anisotropic noise into a single map of the sky. We use this to merge hundreds of individual maps covering the 2008-2018 ACT observing seasons, resulting in by far the deepest ACT maps released so far. We also combine the maps with the full Planck maps, res… ▽ More This paper presents a maximum-likelihood algorithm for combining sky maps with disparate sky coverage, angular resolution and spatially varying anisotropic noise into a single map of the sky. We use this to merge hundreds of individual maps covering the 2008-2018 ACT observing seasons, resulting in by far the deepest ACT maps released so far. We also combine the maps with the full Planck maps, resulting in maps that have the best features of both Planck and ACT: Planck's nearly white noise on intermediate and large angular scales and ACT's high-resolution and sensitivity on small angular scales. The maps cover over 18,000 square degrees, nearly half the full sky, at 100, 150 and 220 GHz. They reveal 4,000 optically-confirmed clusters through the Sunyaev Zel'dovich effect (SZ) and 18,500 point source candidates at $> 5σ$, the largest single collection of SZ clusters and millimeter wave sources to date. The multi-frequency maps provide millimeter images of nearby galaxies and individual Milky Way nebulae, and even clear detections of several nearby stars. Other anticipated uses of these maps include, for example, thermal SZ and kinematic SZ cluster stacking, CMB cluster lensing and galactic dust science. The method itself has negligible bias. However, due to the preliminary nature of some of the component data sets, we caution that these maps should not be used for precision cosmological analysis. The maps are part of ACT DR5, and are available on LAMBDA at https://lambda.gsfc.nasa.gov/product/act/actpol_prod_table.cfm. There is also a web atlas at https://phy-act1.princeton.edu/public/snaess/actpol/dr5/atlas. △ Less

Submitted 17 February, 2021; v1 submitted 14 July, 2020; originally announced July 2020.

Comments: 38 pages, 29 figures, data release on lambda. Published in JCAP

The Atacama Cosmology Telescope: A Measurement of the Cosmic Microwave Background Power Spectra at 98 and 150 GHz

Authors: Steve K. Choi, Matthew Hasselfield, Shuay-Pwu Patty Ho, Brian Koopman, Marius Lungu, Maximilian H. Abitbol, Graeme E. Addison, Peter A. R. Ade, Simone Aiola, David Alonso, Mandana Amiri, Stefania Amodeo, Elio Angile, Jason E. Austermann, Taylor Baildon, Nick Battaglia, James A. Beall, Rachel Bean, Daniel T. Becker, J Richard Bond, Sarah Marie Bruno, Erminia Calabrese, Victoria Calafut, Luis E. Campusano, Felipe Carrero , et al. (114 additional authors not shown)

Abstract: We present the temperature and polarization angular power spectra of the CMB measured by the Atacama Cosmology Telescope (ACT) from 5400 deg$^2$ of the 2013-2016 survey, which covers $>$15000 deg$^2$ at 98 and 150 GHz. For this analysis we adopt a blinding strategy to help avoid confirmation bias and, related to this, show numerous checks for systematic error done before unblinding. Using the like… ▽ More We present the temperature and polarization angular power spectra of the CMB measured by the Atacama Cosmology Telescope (ACT) from 5400 deg$^2$ of the 2013-2016 survey, which covers $>$15000 deg$^2$ at 98 and 150 GHz. For this analysis we adopt a blinding strategy to help avoid confirmation bias and, related to this, show numerous checks for systematic error done before unblinding. Using the likelihood for the cosmological analysis we constrain secondary sources of anisotropy and foreground emission, and derive a "CMB-only" spectrum that extends to $\ell=4000$. At large angular scales, foreground emission at 150 GHz is $\sim$1% of TT and EE within our selected regions and consistent with that found by Planck. Using the same likelihood, we obtain the cosmological parameters for $Λ$CDM for the ACT data alone with a prior on the optical depth of $τ=0.065\pm0.015$. $Λ$CDM is a good fit. The best-fit model has a reduced $χ^2$ of 1.07 (PTE=0.07) with $H_0=67.9\pm1.5$ km/s/Mpc. We show that the lensing BB signal is consistent with $Λ$CDM and limit the celestial EB polarization angle to $ψ_P =-0.07^{\circ}\pm0.09^{\circ}$. We directly cross correlate ACT with Planck and observe generally good agreement but with some discrepancies in TE. All data on which this analysis is based will be publicly released. △ Less

Submitted 23 November, 2020; v1 submitted 14 July, 2020; originally announced July 2020.

Comments: 44 pages, 27 figures, products available on the NASA LAMBDA website, version accepted for publication in JCAP

The Atacama Cosmology Telescope: DR4 Maps and Cosmological Parameters

Authors: Simone Aiola, Erminia Calabrese, Loïc Maurin, Sigurd Naess, Benjamin L. Schmitt, Maximilian H. Abitbol, Graeme E. Addison, Peter A. R. Ade, David Alonso, Mandana Amiri, Stefania Amodeo, Elio Angile, Jason E. Austermann, Taylor Baildon, Nick Battaglia, James A. Beall, Rachel Bean, Daniel T. Becker, J Richard Bond, Sarah Marie Bruno, Victoria Calafut, Luis E. Campusano, Felipe Carrero, Grace E. Chesmore, Hsiao-mei Cho , et al. (116 additional authors not shown)

Abstract: We present new arcminute-resolution maps of the Cosmic Microwave Background temperature and polarization anisotropy from the Atacama Cosmology Telescope, using data taken from 2013-2016 at 98 and 150 GHz. The maps cover more than 17,000 deg$^2$, the deepest 600 deg$^2$ with noise levels below 10 $μ$K-arcmin. We use the power spectrum derived from almost 6,000 deg$^2$ of these maps to constrain cos… ▽ More We present new arcminute-resolution maps of the Cosmic Microwave Background temperature and polarization anisotropy from the Atacama Cosmology Telescope, using data taken from 2013-2016 at 98 and 150 GHz. The maps cover more than 17,000 deg$^2$, the deepest 600 deg$^2$ with noise levels below 10 $μ$K-arcmin. We use the power spectrum derived from almost 6,000 deg$^2$ of these maps to constrain cosmology. The ACT data enable a measurement of the angular scale of features in both the divergence-like polarization and the temperature anisotropy, tracing both the velocity and density at last-scattering. From these one can derive the distance to the last-scattering surface and thus infer the local expansion rate, $H_0$. By combining ACT data with large-scale information from WMAP we measure $H_0 = 67.6 \pm 1.1$ km/s/Mpc, at 68% confidence, in excellent agreement with the independently-measured Planck satellite estimate (from ACT alone we find $H_0 = 67.9 \pm 1.5$ km/s/Mpc). The $Λ$CDM model provides a good fit to the ACT data, and we find no evidence for deviations: both the spatial curvature, and the departure from the standard lensing signal in the spectrum, are zero to within 1$σ$; the number of relativistic species, the primordial Helium fraction, and the running of the spectral index are consistent with $Λ$CDM predictions to within $1.5 - 2.2σ$. We compare ACT, WMAP, and Planck at the parameter level and find good consistency; we investigate how the constraints on the correlated spectral index and baryon density parameters readjust when adding CMB large-scale information that ACT does not measure. The DR4 products presented here will be publicly released on the NASA Legacy Archive for Microwave Background Data Analysis. △ Less

Submitted 3 December, 2020; v1 submitted 14 July, 2020; originally announced July 2020.

Comments: 33 pages, 24 figures, products available on the NASA LAMBDA website, version accepted for publication in JCAP

The Atacama Cosmology Telescope: A CMB lensing mass map over 2100 square degrees of sky and its cross-correlation with BOSS-CMASS galaxies

Authors: Omar Darwish, Mathew S. Madhavacheril, Blake Sherwin, Simone Aiola, Nicholas Battaglia, James A. Beall, Daniel T. Becker, J. Richard Bond, Erminia Calabrese, Steve Choi, Mark J. Devlin, Jo Dunkley, Rolando Dünner, Simone Ferraro, Anna E. Fox, Patricio A. Gallardo, Yilun Guan, Mark Halpern, Dongwon Han, Matthew Hasselfield, J. Colin Hill, Gene C. Hilton, Matt Hilton, Adam D. Hincks, Shuay-Pwu Patty Ho , et al. (28 additional authors not shown)

Abstract: We construct cosmic microwave background lensing mass maps using data from the 2014 and 2015 seasons of observations with the Atacama Cosmology Telescope (ACT). These maps cover 2100 square degrees of sky and overlap with a wide variety of optical surveys. The maps are signal dominated on large scales and have fidelity such that their correlation with the cosmic infrared background is clearly visi… ▽ More We construct cosmic microwave background lensing mass maps using data from the 2014 and 2015 seasons of observations with the Atacama Cosmology Telescope (ACT). These maps cover 2100 square degrees of sky and overlap with a wide variety of optical surveys. The maps are signal dominated on large scales and have fidelity such that their correlation with the cosmic infrared background is clearly visible by eye. We also create lensing maps with thermal Sunyaev-Zel'dovich contamination removed using a novel cleaning procedure that only slightly degrades the lensing signal-to-noise ratio. The cross-spectrum between the cleaned lensing map and the BOSS CMASS galaxy sample is detected at $10$-$σ$ significance, with an amplitude of $A=1.02 \pm 0.10$ relative to the Planck best-fit LCDM cosmological model with fiducial linear galaxy bias. Our measurement lays the foundation for lensing cross-correlation science with current ACT data and beyond. △ Less

Submitted 3 April, 2020; v1 submitted 2 April, 2020; originally announced April 2020.

Comments: 16 pages, 11 figures, lensing map products will be made available on LAMBDA as part of the upcoming ACT data release, v2 corrects author list

Improving Reproducibility in Machine Learning Research (A Report from the NeurIPS 2019 Reproducibility Program)

Authors: Joelle Pineau, Philippe Vincent-Lamarre, Koustuv Sinha, Vincent Larivière, Alina Beygelzimer, Florence d'Alché-Buc, Emily Fox, Hugo Larochelle

Abstract: One of the challenges in machine learning research is to ensure that presented and published results are sound and reliable. Reproducibility, that is obtaining similar results as presented in a paper or talk, using the same code and data (when available), is a necessary step to verify the reliability of research findings. Reproducibility is also an important step to promote open and accessible res… ▽ More One of the challenges in machine learning research is to ensure that presented and published results are sound and reliable. Reproducibility, that is obtaining similar results as presented in a paper or talk, using the same code and data (when available), is a necessary step to verify the reliability of research findings. Reproducibility is also an important step to promote open and accessible research, thereby allowing the scientific community to quickly integrate new findings and convert ideas to practice. Reproducibility also promotes the use of robust experimental workflows, which potentially reduce unintentional errors. In 2019, the Neural Information Processing Systems (NeurIPS) conference, the premier international conference for research in machine learning, introduced a reproducibility program, designed to improve the standards across the community for how we conduct, communicate, and evaluate machine learning research. The program contained three components: a code submission policy, a community-wide reproducibility challenge, and the inclusion of the Machine Learning Reproducibility checklist as part of the paper submission process. In this paper, we describe each of these components, how it was deployed, as well as what we were able to learn from this initiative. △ Less

Submitted 30 December, 2020; v1 submitted 26 March, 2020; originally announced March 2020.

Comments: To appear at JMLR, 16 pages + Appendix

Natural Language Processing Advancements By Deep Learning: A Survey

Authors: Amirsina Torfi, Rouzbeh A. Shirvani, Yaser Keneshloo, Nader Tavaf, Edward A. Fox

Abstract: Natural Language Processing (NLP) helps empower intelligent machines by enhancing a better understanding of the human language for linguistic-based human-computer communication. Recent developments in computational power and the advent of large amounts of linguistic data have heightened the need and demand for automating semantic analysis using data-driven approaches. The utilization of data-drive… ▽ More Natural Language Processing (NLP) helps empower intelligent machines by enhancing a better understanding of the human language for linguistic-based human-computer communication. Recent developments in computational power and the advent of large amounts of linguistic data have heightened the need and demand for automating semantic analysis using data-driven approaches. The utilization of data-driven strategies is pervasive now due to the significant improvements demonstrated through the usage of deep learning methods in areas such as Computer Vision, Automatic Speech Recognition, and in particular, NLP. This survey categorizes and addresses the different aspects and applications of NLP that have benefited from deep learning. It covers core NLP tasks and applications and describes how deep learning methods and models advance these areas. We further analyze and compare different approaches and state-of-the-art models. △ Less

Submitted 27 February, 2021; v1 submitted 2 March, 2020; originally announced March 2020.

The Atacama Cosmology Telescope: Constraints on Cosmic Birefringence

Authors: Toshiya Namikawa, Yilun Guan, Omar Darwish, Blake D. Sherwin, Simone Aiola, Nicholas Battaglia, James A. Beall, Daniel T. Becker, J. Richard Bond, Erminia Calabrese, Grace E. Chesmore, Steve K. Choi, Mark J. Devlin, Joanna Dunkley, Rolando Dünner, Anna E. Fox, Patricio A. Gallardo, Vera Gluscevic, Dongwon Han, Matthew Hasselfield, Gene C. Hilton, Adam D. Hincks, Renée Hložek, Johannes Hubmayr, Kevin Huffenberger , et al. (29 additional authors not shown)

Abstract: We present new constraints on anisotropic birefringence of the cosmic microwave background polarization using two seasons of data from the Atacama Cosmology Telescope covering $456$ square degrees of sky. The birefringence power spectrum, measured using a curved-sky quadratic estimator, is consistent with zero. Our results provide the tightest current constraint on birefringence over a range of an… ▽ More We present new constraints on anisotropic birefringence of the cosmic microwave background polarization using two seasons of data from the Atacama Cosmology Telescope covering $456$ square degrees of sky. The birefringence power spectrum, measured using a curved-sky quadratic estimator, is consistent with zero. Our results provide the tightest current constraint on birefringence over a range of angular scales between $5$ arcminutes and $9$ degrees. We improve previous upper limits on the amplitude of a scale-invariant birefringence power spectrum by a factor of between $2$ and $3$. Assuming a nearly-massless axion field during inflation, our result is equivalent to a $2\,σ$ upper limit on the Chern-Simons coupling constant between axions and photons of $g_{αγ}<4.0\times 10^{-2}/H_I$ where $H_I$ is the inflationary Hubble scale. △ Less

Submitted 21 April, 2020; v1 submitted 28 January, 2020; originally announced January 2020.

Comments: 18 pages, 3 figures, Accepted for publication in PRD

Journal ref: Phys. Rev. D 101, 083527 (2020)

CorGAN: Correlation-Capturing Convolutional Generative Adversarial Networks for Generating Synthetic Healthcare Records

Authors: Amirsina Torfi, Edward A. Fox

Abstract: Deep learning models have demonstrated high-quality performance in areas such as image classification and speech processing. However, creating a deep learning model using electronic health record (EHR) data, requires addressing particular privacy challenges that are unique to researchers in this domain. This matter focuses attention on generating realistic synthetic data while ensuring privacy. In… ▽ More Deep learning models have demonstrated high-quality performance in areas such as image classification and speech processing. However, creating a deep learning model using electronic health record (EHR) data, requires addressing particular privacy challenges that are unique to researchers in this domain. This matter focuses attention on generating realistic synthetic data while ensuring privacy. In this paper, we propose a novel framework called correlation-capturing Generative Adversarial Network (CorGAN), to generate synthetic healthcare records. In CorGAN we utilize Convolutional Neural Networks to capture the correlations between adjacent medical features in the data representation space by combining Convolutional Generative Adversarial Networks and Convolutional Autoencoders. To demonstrate the model fidelity, we show that CorGAN generates synthetic data with performance similar to that of real data in various Machine Learning settings such as classification and prediction. We also give a privacy assessment and report on statistical analysis regarding realistic characteristics of the synthetic data. The software of this work is open-source and is available at: https://github.com/astorfi/cor-gan. △ Less

Submitted 4 March, 2020; v1 submitted 25 January, 2020; originally announced January 2020.

Comments: Accepted to be published in the 33rd International FLAIRS Conference, AI in Healthcare Informatics

Modeling patterns of smartphone usage and their relationship to cognitive health

Authors: Jonas Rauber, Emily B. Fox, Leon A. Gatys

Abstract: The ubiquity of smartphone usage in many people's lives make it a rich source of information about a person's mental and cognitive state. In this work we analyze 12 weeks of phone usage data from 113 older adults, 31 with diagnosed cognitive impairment and 82 without. We develop structured models of users' smartphone interactions to reveal differences in phone usage patterns between people with an… ▽ More The ubiquity of smartphone usage in many people's lives make it a rich source of information about a person's mental and cognitive state. In this work we analyze 12 weeks of phone usage data from 113 older adults, 31 with diagnosed cognitive impairment and 82 without. We develop structured models of users' smartphone interactions to reveal differences in phone usage patterns between people with and without cognitive impairment. In particular, we focus on inferring specific types of phone usage sessions that are predictive of cognitive impairment. Our model achieves an AUROC of 0.79 when discriminating between healthy and symptomatic subjects, and its interpretability enables novel insights into which aspects of phone usage strongly relate with cognitive health in our dataset. △ Less

Submitted 13 November, 2019; originally announced November 2019.

Comments: Machine Learning for Health (ML4H) at NeurIPS 2019 - Extended Abstract

Adaptively Truncating Backpropagation Through Time to Control Gradient Bias

Authors: Christopher Aicher, Nicholas J. Foti, Emily B. Fox

Abstract: Truncated backpropagation through time (TBPTT) is a popular method for learning in recurrent neural networks (RNNs) that saves computation and memory at the cost of bias by truncating backpropagation after a fixed number of lags. In practice, choosing the optimal truncation length is difficult: TBPTT will not converge if the truncation length is too small, or will converge slowly if it is too larg… ▽ More Truncated backpropagation through time (TBPTT) is a popular method for learning in recurrent neural networks (RNNs) that saves computation and memory at the cost of bias by truncating backpropagation after a fixed number of lags. In practice, choosing the optimal truncation length is difficult: TBPTT will not converge if the truncation length is too small, or will converge slowly if it is too large. We propose an adaptive TBPTT scheme that converts the problem from choosing a temporal lag to one of choosing a tolerable amount of gradient bias. For many realistic RNNs, the TBPTT gradients decay geometrically in expectation for large lags; under this condition, we can control the bias by varying the truncation length adaptively. For RNNs with smooth activation functions, we prove that this bias controls the convergence rate of SGD with biased gradients for our non-convex loss. Using this theory, we develop a practical method for adaptively estimating the truncation length during training. We evaluate our adaptive TBPTT method on synthetic data and language modeling tasks and find that our adaptive TBPTT ameliorates the computational pitfalls of fixed TBPTT. △ Less

Submitted 1 July, 2019; v1 submitted 17 May, 2019; originally announced May 2019.

Tunable Correlated Chern Insulator and Ferromagnetism in Trilayer Graphene/Boron Nitride Moiré Superlattice

Authors: Guorui Chen, Aaron L. Sharpe, Eli J. Fox, Ya-Hui Zhang, Shaoxin Wang, Lili Jiang, Bosai Lyu, Hongyuan Li, Kenji Watanabe, Takashi Taniguchi, Zhiwen Shi, T. Senthil, David Goldhaber-Gordon, Yuanbo Zhang, Feng Wang

Abstract: Studies on two-dimensional electron systems in a strong magnetic field first revealed the quantum Hall (QH) effect, a topological state of matter featuring a finite Chern number (C) and chiral edge states. Haldane later theorized that Chern insulators with integer QH effects could appear in lattice models with complex hopping parameters even at zero magnetic field. The ABC-trilayer graphene/hexago… ▽ More Studies on two-dimensional electron systems in a strong magnetic field first revealed the quantum Hall (QH) effect, a topological state of matter featuring a finite Chern number (C) and chiral edge states. Haldane later theorized that Chern insulators with integer QH effects could appear in lattice models with complex hopping parameters even at zero magnetic field. The ABC-trilayer graphene/hexagonal boron nitride (TLG/hBN) moiré superlattice provides an attractive platform to explore Chern insulators because it features nearly flat moiré minibands with a valley-dependent electrically tunable Chern number. Here we report the experimental observation of a correlated Chern insulator in a TLG/hBN moiré superlattice. We show that reversing the direction of the applied vertical electric field switches TLG/hBN's moiré minibands between zero and finite Chern numbers, as revealed by dramatic changes in magneto-transport behavior. For topological hole minibands tuned to have a finite Chern number, we focus on 1/4 filling, corresponding to one hole per moiré unit cell. The Hall resistance is well quantized at h/2e2, i.e. C = 2, for |B| > 0.4 T. The correlated Chern insulator is ferromagnetic, exhibiting significant magnetic hysteresis and a large anomalous Hall signal at zero magnetic field. Our discovery of a C = 2 Chern insulator at zero magnetic field should open up exciting opportunities for discovering novel correlated topological states, possibly with novel topological excitations, in nearly flat and topologically nontrivial moiré minibands. △ Less

Submitted 16 May, 2019; originally announced May 2019.

Comments: 16 pages, 4 figures, and 2 extended figures

Stochastic Gradient MCMC for Nonlinear State Space Models

Authors: Christopher Aicher, Srshti Putcha, Christopher Nemeth, Paul Fearnhead, Emily B. Fox

Abstract: State space models (SSMs) provide a flexible framework for modeling complex time series via a latent stochastic process. Inference for nonlinear, non-Gaussian SSMs is often tackled with particle methods that do not scale well to long time series. The challenge is two-fold: not only do computations scale linearly with time, as in the linear case, but particle filters additionally suffer from increa… ▽ More State space models (SSMs) provide a flexible framework for modeling complex time series via a latent stochastic process. Inference for nonlinear, non-Gaussian SSMs is often tackled with particle methods that do not scale well to long time series. The challenge is two-fold: not only do computations scale linearly with time, as in the linear case, but particle filters additionally suffer from increasing particle degeneracy with longer series. Stochastic gradient MCMC methods have been developed to scale Bayesian inference for finite-state hidden Markov models and linear SSMs using buffered stochastic gradient estimates to account for temporal dependencies. We extend these stochastic gradient estimators to nonlinear SSMs using particle methods. We present error bounds that account for both buffering error and particle error in the case of nonlinear SSMs that are log-concave in the latent process. We evaluate our proposed particle buffered stochastic gradient using stochastic gradient MCMC for inference on both long sequential synthetic and minute-resolution financial returns data, demonstrating the importance of this class of methods. △ Less

Submitted 16 July, 2023; v1 submitted 29 January, 2019; originally announced January 2019.

Comments: To appear in Bayesian Analysis

Signatures of Gate-Tunable Superconductivity in Trilayer Graphene/Boron Nitride Moiré Superlattice

Authors: Guorui Chen, Aaron L. Sharpe, Patrick Gallagher, Ilan T. Rosen, Eli Fox, Lili Jiang, Bosai Lyu, Hongyuan Li, Kenji Watanabe, Takashi Taniguchi, Jeil Jung, Zhiwen Shi, David Goldhaber-Gordon, Yuanbo Zhang, Feng Wang

Abstract: Understanding the mechanism of high temperature (high Tc) superconductivity is a central problem in condensed matter physics. It is often speculated that high Tc superconductivity arises from a doped Mott insulator as described by the Hubbard model. An exact solution of the Hubbard model, however, is extremely challenging due to the strong electron-electron correlation. Therefore, it is highly des… ▽ More Understanding the mechanism of high temperature (high Tc) superconductivity is a central problem in condensed matter physics. It is often speculated that high Tc superconductivity arises from a doped Mott insulator as described by the Hubbard model. An exact solution of the Hubbard model, however, is extremely challenging due to the strong electron-electron correlation. Therefore, it is highly desirable to experimentally study a model Hubbard system in which the unconventional superconductivity can be continuously tuned by varying the Hubbard parameters. Here we report signatures of tunable superconductivity in ABC-trilayer graphene (TLG) / boron nitride (hBN) moiré superlattice. Unlike "magic angle" twisted bilayer graphene, theoretical calculations show that under a vertical displacement field the ABC-TLG/hBN heterostructure features an isolated flat valence miniband associated with a Hubbard model on a triangular superlattice. Upon applying such a displacement field we find experimentally that the ABC-TLG/hBN superlattice displays Mott insulating states below 20 Kelvin at 1/4 and 1/2 fillings, corresponding to 1 and 2 holes per unit cell, respectively. Upon further cooling, signatures of superconducting domes emerge below 1 kelvin for the electron- and hole-doped sides of the 1/4 filling Mott state. The electronic behavior in the TLG/hBN superlattice is expected to depend sensitively on the interplay between the electron-electron interaction and the miniband bandwidth, which can be tuned continuously with the displacement field D. By simply varying the D field, we demonstrate transitions from the candidate superconductor to Mott insulator and metallic phases. Our study shows that TLG/hBN heterostructures offer an attractive model system to explore rich correlated behavior emerging in the tunable triangular Hubbard model. △ Less

Submitted 14 January, 2019; originally announced January 2019.

Comments: 14 pages, 4 figures

Journal ref: Nature 572, 215-219 (2019)

Emergent ferromagnetism near three-quarters filling in twisted bilayer graphene

Authors: Aaron L. Sharpe, Eli J. Fox, Arthur W. Barnard, Joe Finney, Kenji Watanabe, Takashi Taniguchi, M. A. Kastner, David Goldhaber-Gordon

Abstract: When two sheets of graphene are stacked at a small twist angle, the resulting flat superlattice minibands are expected to strongly enhance electron-electron interactions. Here we present evidence that near three-quarters ($3/4$) filling of the conduction miniband these enhanced interactions drive the twisted bilayer graphene into a ferromagnetic state. We observe emergent ferromagnetic hysteresis,… ▽ More When two sheets of graphene are stacked at a small twist angle, the resulting flat superlattice minibands are expected to strongly enhance electron-electron interactions. Here we present evidence that near three-quarters ($3/4$) filling of the conduction miniband these enhanced interactions drive the twisted bilayer graphene into a ferromagnetic state. We observe emergent ferromagnetic hysteresis, with a giant anomalous Hall (AH) effect as large as $10.4\ \mathrm{kΩ}$ and signs of chiral edge states in a narrow density range around an apparent insulating state at $3/4$. Surprisingly, the magnetization of the sample can be reversed by applying a small DC current. Although the AH resistance is not quantized and dissipation is significant, we suggest that the system is an incipient Chern insulator. △ Less

Submitted 11 January, 2019; originally announced January 2019.

Comments: 18 pages, 4 figures

Journal ref: Science 365, 605-608 (2019)

Comparing Spatial Regression to Random Forests for Large Environmental Data Sets

Authors: Eric W. Fox, Jay M. Ver Hoef, Anthony R. Olsen

Abstract: Environmental data may be "large" due to number of records, number of covariates, or both. Random forests has a reputation for good predictive performance when using many covariates with nonlinear relationships, whereas spatial regression, when using reduced rank methods, has a reputation for good predictive performance when using many records that are spatially autocorrelated. In this study, we c… ▽ More Environmental data may be "large" due to number of records, number of covariates, or both. Random forests has a reputation for good predictive performance when using many covariates with nonlinear relationships, whereas spatial regression, when using reduced rank methods, has a reputation for good predictive performance when using many records that are spatially autocorrelated. In this study, we compare these two techniques using a data set containing the macroinvertebrate multimetric index (MMI) at 1859 stream sites with over 200 landscape covariates. A primary application is mapping MMI predictions and prediction errors at 1.1 million perennial stream reaches across the conterminous United States. For the spatial regression model, we develop a novel transformation procedure that estimates Box-Cox transformations to linearize covariate relationships and handles possibly zero-inflated covariates. We find that the spatial regression model with transformations, and a subsequent selection of significant covariates, has cross-validation performance slightly better than random forests. We also find that prediction interval coverage is close to nominal for each method, but that spatial regression prediction intervals tend to be narrower and have less variability than quantile regression forest prediction intervals. A simulation study is used to generalize results and clarify advantages of each modeling approach. △ Less

Submitted 26 December, 2018; originally announced December 2018.

A Hybrid Model for Role-related User Classification on Twitter

Authors: Liuqing Li, Ziqian Song, Xuan Zhang, Edward A. Fox

Abstract: To aid a variety of research studies, we propose TWIROLE, a hybrid model for role-related user classification on Twitter, which detects male-related, female-related, and brand-related (i.e., organization or institution) users. TWIROLE leverages features from tweet contents, user profiles, and profile images, and then applies our hybrid model to identify a user's role. To evaluate it, we used two e… ▽ More To aid a variety of research studies, we propose TWIROLE, a hybrid model for role-related user classification on Twitter, which detects male-related, female-related, and brand-related (i.e., organization or institution) users. TWIROLE leverages features from tweet contents, user profiles, and profile images, and then applies our hybrid model to identify a user's role. To evaluate it, we used two existing large datasets about Twitter users, and conducted both intra- and inter-comparison experiments. TWIROLE outperforms existing methods and obtains more balanced results over the several roles. We also confirm that user names and profile images are good indicators for this task. Our research extends prior work that does not consider brand-related users, and is an aid to future evaluation efforts relative to investigations that rely upon self-labeled datasets. △ Less

Submitted 26 November, 2018; originally announced November 2018.

First results from the LUCID-Timepix spacecraft payload onboard the TechDemoSat-1 satellite in Low Earth Orbit

Authors: Will Furnell, Abhishek Shenoy, Elliot Fox, Peter Hatfield

Abstract: The Langton Ultimate Cosmic ray Intensity Detector (LUCID) is a payload onboard the satellite TechDemoSat-1, used to study the radiation environment in Low Earth Orbit ($\sim$635km). LUCID operated from 2014 to 2017, collecting over 2.1 million frames of radiation data from its five Timepix detectors on board. LUCID is one of the first uses of the Timepix detector technology in open space, with th… ▽ More The Langton Ultimate Cosmic ray Intensity Detector (LUCID) is a payload onboard the satellite TechDemoSat-1, used to study the radiation environment in Low Earth Orbit ($\sim$635km). LUCID operated from 2014 to 2017, collecting over 2.1 million frames of radiation data from its five Timepix detectors on board. LUCID is one of the first uses of the Timepix detector technology in open space, with the data providing useful insight into the performance of this technology in new environments. It provides high-sensitivity imaging measurements of the mixed radiation field, with a wide dynamic range in terms of spectral response, particle type and direction. The data has been analysed using computing resources provided by GridPP, with a new machine learning algorithm that uses the Tensorflow framework. This algorithm provides a new approach to processing Medipix data, using a training set of human labelled tracks, providing greater particle classification accuracy than other algorithms. For managing the LUCID data, we have developed an online platform called Timepix Analysis Platform at School (TAPAS). This provides a swift and simple way for users to analyse data that they collect using Timepix detectors from both LUCID and other experiments. We also present some possible future uses of the LUCID data and Medipix detectors in space. △ Less

Submitted 30 October, 2018; originally announced October 2018.

Comments: Accepted for publication in Advances in Space Research

Stochastic Gradient MCMC for State Space Models

Authors: Christopher Aicher, Yi-An Ma, Nicholas J. Foti, Emily B. Fox

Abstract: State space models (SSMs) are a flexible approach to modeling complex time series. However, inference in SSMs is often computationally prohibitive for long time series. Stochastic gradient MCMC (SGMCMC) is a popular method for scalable Bayesian inference for large independent data. Unfortunately when applied to dependent data, such as in SSMs, SGMCMC's stochastic gradient estimates are biased as t… ▽ More State space models (SSMs) are a flexible approach to modeling complex time series. However, inference in SSMs is often computationally prohibitive for long time series. Stochastic gradient MCMC (SGMCMC) is a popular method for scalable Bayesian inference for large independent data. Unfortunately when applied to dependent data, such as in SSMs, SGMCMC's stochastic gradient estimates are biased as they break crucial temporal dependencies. To alleviate this, we propose stochastic gradient estimators that control this bias by performing additional computation in a `buffer' to reduce breaking dependencies. Furthermore, we derive error bounds for this bias and show a geometric decay under mild conditions. Using these estimators, we develop novel SGMCMC samplers for discrete, continuous and mixed-type SSMs with analytic message passing. Our experiments on real and synthetic data demonstrate the effectiveness of our SGMCMC algorithms compared to batch MCMC, allowing us to scale inference to long time series with millions of time points. △ Less

Submitted 9 July, 2019; v1 submitted 22 October, 2018; originally announced October 2018.

Approximate Collapsed Gibbs Clustering with Expectation Propagation

Authors: Christopher Aicher, Emily B. Fox

Abstract: We develop a framework for approximating collapsed Gibbs sampling in generative latent variable cluster models. Collapsed Gibbs is a popular MCMC method, which integrates out variables in the posterior to improve mixing. Unfortunately for many complex models, integrating out these variables is either analytically or computationally intractable. We efficiently approximate the necessary collapsed Gi… ▽ More We develop a framework for approximating collapsed Gibbs sampling in generative latent variable cluster models. Collapsed Gibbs is a popular MCMC method, which integrates out variables in the posterior to improve mixing. Unfortunately for many complex models, integrating out these variables is either analytically or computationally intractable. We efficiently approximate the necessary collapsed Gibbs integrals by borrowing ideas from expectation propagation. We present two case studies where exact collapsed Gibbs sampling is intractable: mixtures of Student-t's and time series clustering. Our experiments on real and synthetic data show that our approximate sampler enables a runtime-accuracy tradeoff in sampling these types of models, providing results with competitive accuracy much more rapidly than the naive Gibbs samplers one would otherwise rely on in these scenarios. △ Less

Submitted 19 July, 2018; originally announced July 2018.

Disentangled VAE Representations for Multi-Aspect and Missing Data

Authors: Samuel K. Ainsworth, Nicholas J. Foti, Emily B. Fox

Abstract: Many problems in machine learning and related application areas are fundamentally variants of conditional modeling and sampling across multi-aspect data, either multi-view, multi-modal, or simply multi-group. For example, sampling from the distribution of English sentences conditioned on a given French sentence or sampling audio waveforms conditioned on a given piece of text. Central to many of th… ▽ More Many problems in machine learning and related application areas are fundamentally variants of conditional modeling and sampling across multi-aspect data, either multi-view, multi-modal, or simply multi-group. For example, sampling from the distribution of English sentences conditioned on a given French sentence or sampling audio waveforms conditioned on a given piece of text. Central to many of these problems is the issue of missing data: we can observe many English, French, or German sentences individually but only occasionally do we have data for a sentence pair. Motivated by these applications and inspired by recent progress in variational autoencoders for grouped data, we develop factVAE, a deep generative model capable of handling multi-aspect data, robust to missing observations, and with a prior that encourages disentanglement between the groups and the latent dimensions. The effectiveness of factVAE is demonstrated on a variety of rich real-world datasets, including motion capture poses and pictures of faces captured from varying poses and perspectives. △ Less

Submitted 23 June, 2018; originally announced June 2018.

Large-Scale Stochastic Sampling from the Probability Simplex

Authors: Jack Baker, Paul Fearnhead, Emily B Fox, Christopher Nemeth

Abstract: Stochastic gradient Markov chain Monte Carlo (SGMCMC) has become a popular method for scalable Bayesian inference. These methods are based on sampling a discrete-time approximation to a continuous time process, such as the Langevin diffusion. When applied to distributions defined on a constrained space the time-discretization error can dominate when we are near the boundary of the space. We demons… ▽ More Stochastic gradient Markov chain Monte Carlo (SGMCMC) has become a popular method for scalable Bayesian inference. These methods are based on sampling a discrete-time approximation to a continuous time process, such as the Langevin diffusion. When applied to distributions defined on a constrained space the time-discretization error can dominate when we are near the boundary of the space. We demonstrate that because of this, current SGMCMC methods for the simplex struggle with sparse simplex spaces; when many of the components are close to zero. Unfortunately, many popular large-scale Bayesian models, such as network or topic models, require inference on sparse simplex spaces. To avoid the biases caused by this discretization error, we propose the stochastic Cox-Ingersoll-Ross process (SCIR), which removes all discretization error and we prove that samples from the SCIR process are asymptotically unbiased. We discuss how this idea can be extended to target other constrained spaces. Use of the SCIR process within a SGMCMC algorithm is shown to give substantially better performance for a topic model and a Dirichlet process mixture model than existing SGMCMC approaches. △ Less

Submitted 26 October, 2018; v1 submitted 19 June, 2018; originally announced June 2018.

Comments: Accepted to Advances in Neural Information Processing Systems (2018)

Interpretable VAEs for nonlinear group factor analysis

Authors: Samuel Ainsworth, Nicholas Foti, Adrian KC Lee, Emily Fox

Abstract: Deep generative models have recently yielded encouraging results in producing subjectively realistic samples of complex data. Far less attention has been paid to making these generative models interpretable. In many scenarios, ranging from scientific applications to finance, the observed variables have a natural grouping. It is often of interest to understand systems of interaction amongst these g… ▽ More Deep generative models have recently yielded encouraging results in producing subjectively realistic samples of complex data. Far less attention has been paid to making these generative models interpretable. In many scenarios, ranging from scientific applications to finance, the observed variables have a natural grouping. It is often of interest to understand systems of interaction amongst these groups, and latent factor models (LFMs) are an attractive approach. However, traditional LFMs are limited by assuming a linear correlation structure. We present an output interpretable VAE (oi-VAE) for grouped data that models complex, nonlinear latent-to-observed relationships. We combine a structured VAE comprised of group-specific generators with a sparsity-inducing prior. We demonstrate that oi-VAE yields meaningful notions of interpretability in the analysis of motion capture and MEG data. We further show that in these situations, the regularization inherent to oi-VAE can actually lead to improved generalization and learned generative processes. △ Less

Submitted 16 February, 2018; originally announced February 2018.

Neural Granger Causality

Authors: Alex Tank, Ian Covert, Nicholas Foti, Ali Shojaie, Emily Fox

Abstract: While most classical approaches to Granger causality detection assume linear dynamics, many interactions in real-world applications, like neuroscience and genomics, are inherently nonlinear. In these cases, using linear models may lead to inconsistent estimation of Granger causal interactions. We propose a class of nonlinear methods by applying structured multilayer perceptrons (MLPs) or recurrent… ▽ More While most classical approaches to Granger causality detection assume linear dynamics, many interactions in real-world applications, like neuroscience and genomics, are inherently nonlinear. In these cases, using linear models may lead to inconsistent estimation of Granger causal interactions. We propose a class of nonlinear methods by applying structured multilayer perceptrons (MLPs) or recurrent neural networks (RNNs) combined with sparsity-inducing penalties on the weights. By encouraging specific sets of weights to be zero--in particular, through the use of convex group-lasso penalties--we can extract the Granger causal structure. To further contrast with traditional approaches, our framework naturally enables us to efficiently capture long-range dependencies between series either via our RNNs or through an automatic lag selection in the MLP. We show that our neural Granger causality methods outperform state-of-the-art nonlinear Granger causality methods on the DREAM3 challenge data. This data consists of nonlinear gene expression and regulation time courses with only a limited number of time points. The successes we show in this challenging dataset provide a powerful example of how deep learning can be useful in cases that go beyond prediction on large datasets. We likewise illustrate our methods in detecting nonlinear interactions in a human motion capture dataset. △ Less

Submitted 13 March, 2021; v1 submitted 16 February, 2018; originally announced February 2018.

Comments: IEEE TPAMI accepted version

An Efficient ADMM Algorithm for Structural Break Detection in Multivariate Time Series

Authors: Alex Tank, Emily B. Fox, Ali Shojaie

Abstract: We present an efficient alternating direction method of multipliers (ADMM) algorithm for segmenting a multivariate non-stationary time series with structural breaks into stationary regions. We draw from recent work where the series is assumed to follow a vector autoregressive model within segments and a convex estimation procedure may be formulated using group fused lasso penalties. Our ADMM appro… ▽ More We present an efficient alternating direction method of multipliers (ADMM) algorithm for segmenting a multivariate non-stationary time series with structural breaks into stationary regions. We draw from recent work where the series is assumed to follow a vector autoregressive model within segments and a convex estimation procedure may be formulated using group fused lasso penalties. Our ADMM approach first splits the convex problem into a global quadratic program and a simple group lasso proximal update. We show that the global problem may be parallelized over rows of the time dependent transition matrices and furthermore that each subproblem may be rewritten in a form identical to the log-likelihood of a Gaussian state space model. Consequently, we develop a Kalman smoothing algorithm to solve the global update in time linear in the length of the series. △ Less

Submitted 25 June, 2018; v1 submitted 22 November, 2017; originally announced November 2017.

Comments: Accepted to the NIPS Time Series Workshop 2017

An Interpretable and Sparse Neural Network Model for Nonlinear Granger Causality Discovery

Authors: Alex Tank, Ian Cover, Nicholas J. Foti, Ali Shojaie, Emily B. Fox

Abstract: While most classical approaches to Granger causality detection repose upon linear time series assumptions, many interactions in neuroscience and economics applications are nonlinear. We develop an approach to nonlinear Granger causality detection using multilayer perceptrons where the input to the network is the past time lags of all series and the output is the future value of a single series. A… ▽ More While most classical approaches to Granger causality detection repose upon linear time series assumptions, many interactions in neuroscience and economics applications are nonlinear. We develop an approach to nonlinear Granger causality detection using multilayer perceptrons where the input to the network is the past time lags of all series and the output is the future value of a single series. A sufficient condition for Granger non-causality in this setting is that all of the outgoing weights of the input data, the past lags of a series, to the first hidden layer are zero. For estimation, we utilize a group lasso penalty to shrink groups of input weights to zero. We also propose a hierarchical penalty for simultaneous Granger causality and lag estimation. We validate our approach on simulated data from both a sparse linear autoregressive model and the sparse and nonlinear Lorenz-96 model. △ Less

Submitted 25 June, 2018; v1 submitted 22 November, 2017; originally announced November 2017.

Comments: Accepted to the NIPS Time Series Workshop 2017

A Unified Framework for Long Range and Cold Start Forecasting of Seasonal Profiles in Time Series

Authors: Christopher Xie, Alex Tank, Alec Greaves-Tunnell, Emily Fox

Abstract: Providing long-range forecasts is a fundamental challenge in time series modeling, which is only compounded by the challenge of having to form such forecasts when a time series has never previously been observed. The latter challenge is the time series version of the cold-start problem seen in recommender systems which, to our knowledge, has not been addressed in previous work. A similar problem o… ▽ More Providing long-range forecasts is a fundamental challenge in time series modeling, which is only compounded by the challenge of having to form such forecasts when a time series has never previously been observed. The latter challenge is the time series version of the cold-start problem seen in recommender systems which, to our knowledge, has not been addressed in previous work. A similar problem occurs when a long range forecast is required after only observing a small number of time points --- a warm start forecast. With these aims in mind, we focus on forecasting seasonal profiles---or baseline demand---for periods on the order of a year in three cases: the long range case with multiple previously observed seasonal profiles, the cold start case with no previous observed seasonal profiles, and the warm start case with only a single partially observed profile. Classical time series approaches that perform iterated step-ahead forecasts based on previous observations struggle to provide accurate long range predictions; in settings with little to no observed data, such approaches are simply not applicable. Instead, we present a straightforward framework which combines ideas from high-dimensional regression and matrix factorization on a carefully constructed data matrix. Key to our formulation and resulting performance is leveraging (1) repeated patterns over fixed periods of time and across series, and (2) metadata associated with the individual series; without this additional data, the cold-start/warm-start problems are nearly impossible to solve. We demonstrate that our framework can accurately forecast an array of seasonal profiles on multiple large scale datasets. △ Less

Submitted 26 August, 2018; v1 submitted 23 October, 2017; originally announced October 2017.

Part-per-million quantization and current-induced breakdown of the quantum anomalous Hall effect

Authors: E. J. Fox, I. T. Rosen, Yanfei Yang, George R. Jones, Randolph E. Elmquist, Xufeng Kou, Lei Pan, Kang L. Wang, D. Goldhaber-Gordon

Abstract: In the quantum anomalous Hall effect, quantized Hall resistance and vanishing longitudinal resistivity are predicted to result from the presence of dissipationless, chiral edge states and an insulating 2D bulk, without requiring an external magnetic field. Here, we explore the potential of this effect in magnetic topological insulator thin films for metrological applications. Using a cryogenic cur… ▽ More In the quantum anomalous Hall effect, quantized Hall resistance and vanishing longitudinal resistivity are predicted to result from the presence of dissipationless, chiral edge states and an insulating 2D bulk, without requiring an external magnetic field. Here, we explore the potential of this effect in magnetic topological insulator thin films for metrological applications. Using a cryogenic current comparator system, we measure quantization of the Hall resistance to within one part per million and longitudinal resistivity under 10 m$Ω$ per square at zero magnetic field. Increasing the current density past a critical value leads to a breakdown of the quantized, low-dissipation state, which we attribute to electron heating in bulk current flow. We further investigate the pre-breakdown regime by measuring transport dependence on temperature, current, and geometry, and find evidence for bulk dissipation, including thermal activation and possible variable-range hopping. △ Less

Submitted 4 October, 2017; originally announced October 2017.

Comments: 9 pages, 4 figures, with 7 pages of supplementary information

Journal ref: Phys. Rev. B 98, 075145 (2018)

sgmcmc: An R Package for Stochastic Gradient Markov Chain Monte Carlo

Authors: Jack Baker, Paul Fearnhead, Emily B. Fox, Christopher Nemeth

Abstract: This paper introduces the R package sgmcmc; which can be used for Bayesian inference on problems with large datasets using stochastic gradient Markov chain Monte Carlo (SGMCMC). Traditional Markov chain Monte Carlo (MCMC) methods, such as Metropolis-Hastings, are known to run prohibitively slowly as the dataset size increases. SGMCMC solves this issue by only using a subset of data at each iterati… ▽ More This paper introduces the R package sgmcmc; which can be used for Bayesian inference on problems with large datasets using stochastic gradient Markov chain Monte Carlo (SGMCMC). Traditional Markov chain Monte Carlo (MCMC) methods, such as Metropolis-Hastings, are known to run prohibitively slowly as the dataset size increases. SGMCMC solves this issue by only using a subset of data at each iteration. SGMCMC requires calculating gradients of the log likelihood and log priors, which can be time consuming and error prone to perform by hand. The sgmcmc package calculates these gradients itself using automatic differentiation, making the implementation of these methods much easier. To do this, the package uses the software library TensorFlow, which has a variety of statistical distributions and mathematical operations as standard, meaning a wide class of models can be built using this framework. SGMCMC has become widely adopted in the machine learning literature, but less so in the statistics community. We believe this may be partly due to lack of software; this package aims to bridge this gap. △ Less

Submitted 13 April, 2018; v1 submitted 2 October, 2017; originally announced October 2017.

Dynamics of homelessness in urban America

Authors: Chris Glynn, Emily B. Fox

Abstract: The relationship between housing costs and homelessness has important implications for the way that city and county governments respond to increasing homeless populations. Though many analyses in the public policy literature have examined inter-community variation in homelessness rates to identify causal mechanisms of homelessness (Byrne et al., 2013; Lee et al., 2003; Fargo et al., 2013), few stu… ▽ More The relationship between housing costs and homelessness has important implications for the way that city and county governments respond to increasing homeless populations. Though many analyses in the public policy literature have examined inter-community variation in homelessness rates to identify causal mechanisms of homelessness (Byrne et al., 2013; Lee et al., 2003; Fargo et al., 2013), few studies have examined time-varying homeless counts within the same community (McCandless et al., 2016). To examine trends in homeless population counts in the 25 largest U.S. metropolitan areas, we develop a dynamic Bayesian hierarchical model for time-varying homeless count data. Particular care is given to modeling uncertainty in the homeless count generating and measurement processes, and a critical distinction is made between the counted number of homeless and the true size of the homeless population. For each metro under study, we investigate the relationship between increases in the Zillow Rent Index and increases in the homeless population. Sensitivity of inference to potential improvements in the accuracy of point-in-time counts is explored, and evidence is presented that the inferred increase in the rate of homelessness from 2011-2016 depends on prior beliefs about the accuracy of homeless counts. A main finding of the study is that the relationship between homelessness and rental costs is strongest in New York, Los Angeles, Washington, D.C., and Seattle. △ Less

Submitted 28 July, 2017; originally announced July 2017.

Comments: 54 pages, 36 figures

Chiral transport along magnetic domain walls in the quantum anomalous Hall effect

Authors: I. T. Rosen, E. J. Fox, Xufeng Kou, Lei Pan, Kang L. Wang, D. Goldhaber-Gordon

Abstract: The recent prediction, and subsequent discovery, of the quantum anomalous Hall (QAH) effect in thin films of the three-dimensional ferromagnetic topological insulator (MTI) (Cr$_y$Bi$_x$Sb$_{1-x-y}$)$_2$Te$_3$ has opened new possibilities for chiral-edge-state-based devices in zero external magnetic field. Like the $ν=1$ quantum Hall system, the QAH system is predicted to have a single chiral edge… ▽ More The recent prediction, and subsequent discovery, of the quantum anomalous Hall (QAH) effect in thin films of the three-dimensional ferromagnetic topological insulator (MTI) (Cr$_y$Bi$_x$Sb$_{1-x-y}$)$_2$Te$_3$ has opened new possibilities for chiral-edge-state-based devices in zero external magnetic field. Like the $ν=1$ quantum Hall system, the QAH system is predicted to have a single chiral edge mode circulating along the boundary of the film. Backscattering of the chiral edge mode should be suppressed, as recently verified by the observation of well-quantized Hall resistivities $ρ_{yx} = \pm h/e^2$, along with longitudinal resistivities as low as a few ohms. Dissipationless 1D conduction is also expected along magnetic domain walls. Here, we intentionally create a magnetic domain wall in a MTI and study electrical transport along the domain wall. We present the first observation of chiral transport along domain walls, in agreement with theoretical predictions. We present further evidence that two modes equilibrate and co-propagate along the length of the domain wall. △ Less

Submitted 26 July, 2017; originally announced July 2017.

Comments: 5 pages, 3 figures, with 8 pages of supplementary information

Journal ref: npj Quantum Materials 2, 69 (2017)

Control Variates for Stochastic Gradient MCMC

Authors: Jack Baker, Paul Fearnhead, Emily B. Fox, Christopher Nemeth

Abstract: It is well known that Markov chain Monte Carlo (MCMC) methods scale poorly with dataset size. A popular class of methods for solving this issue is stochastic gradient MCMC. These methods use a noisy estimate of the gradient of the log posterior, which reduces the per iteration computational cost of the algorithm. Despite this, there are a number of results suggesting that stochastic gradient Lange… ▽ More It is well known that Markov chain Monte Carlo (MCMC) methods scale poorly with dataset size. A popular class of methods for solving this issue is stochastic gradient MCMC. These methods use a noisy estimate of the gradient of the log posterior, which reduces the per iteration computational cost of the algorithm. Despite this, there are a number of results suggesting that stochastic gradient Langevin dynamics (SGLD), probably the most popular of these methods, still has computational cost proportional to the dataset size. We suggest an alternative log posterior gradient estimate for stochastic gradient MCMC, which uses control variates to reduce the variance. We analyse SGLD using this gradient estimate, and show that, under log-concavity assumptions on the target distribution, the computational cost required for a given level of accuracy is independent of the dataset size. Next we show that a different control variate technique, known as zero variance control variates can be applied to SGMCMC algorithms for free. This post-processing step improves the inference of the algorithm by reducing the variance of the MCMC output. Zero variance control variates rely on the gradient of the log posterior; we explore how the variance reduction is affected by replacing this with the noisy gradient estimate calculated by SGMCMC. △ Less

Submitted 14 December, 2017; v1 submitted 16 June, 2017; originally announced June 2017.

Stochastic Gradient MCMC Methods for Hidden Markov Models

Authors: Yi-An Ma, Nicholas J. Foti, Emily B. Fox

Abstract: Stochastic gradient MCMC (SG-MCMC) algorithms have proven useful in scaling Bayesian inference to large datasets under an assumption of i.i.d data. We instead develop an SG-MCMC algorithm to learn the parameters of hidden Markov models (HMMs) for time-dependent data. There are two challenges to applying SG-MCMC in this setting: The latent discrete states, and needing to break dependencies when con… ▽ More Stochastic gradient MCMC (SG-MCMC) algorithms have proven useful in scaling Bayesian inference to large datasets under an assumption of i.i.d data. We instead develop an SG-MCMC algorithm to learn the parameters of hidden Markov models (HMMs) for time-dependent data. There are two challenges to applying SG-MCMC in this setting: The latent discrete states, and needing to break dependencies when considering minibatches. We consider a marginal likelihood representation of the HMM and propose an algorithm that harnesses the inherent memory decay of the process. We demonstrate the effectiveness of our algorithm on synthetic experiments and an ion channel recording data, with runtimes significantly outperforming batch MCMC. △ Less

Submitted 14 June, 2017; originally announced June 2017.

Granger Causality Networks for Categorical Time Series

Authors: Alex Tank, Emily B. Fox, Ali Shojaie

Abstract: We present a new framework for learning Granger causality networks for multivariate categorical time series, based on the mixture transition distribution (MTD) model. Traditionally, MTD is plagued by a nonconvex objective, non-identifiability, and presence of many local optima. To circumvent these problems, we recast inference in the MTD as a convex problem. The new formulation facilitates the app… ▽ More We present a new framework for learning Granger causality networks for multivariate categorical time series, based on the mixture transition distribution (MTD) model. Traditionally, MTD is plagued by a nonconvex objective, non-identifiability, and presence of many local optima. To circumvent these problems, we recast inference in the MTD as a convex problem. The new formulation facilitates the application of MTD to high-dimensional multivariate time series. As a baseline, we also formulate a multi-output logistic autoregressive model (mLTD), which while a straightforward extension of autoregressive Bernoulli generalized linear models, has not been previously applied to the analysis of multivariate categorial time series. We develop novel identifiability conditions of the MTD model and compare them to those for mLTD. We further devise novel and efficient optimization algorithm for the MTD based on the new convex formulation, and compare the MTD and mLTD in both simulated and real data experiments. Our approach simultaneously provides a comparison of methods for network inference in categorical time series and opens the door to modern, regularized inference with the MTD model. △ Less

Submitted 8 June, 2017; originally announced June 2017.

Identifiability and Estimation of Structural Vector Autoregressive Models for Subsampled and Mixed Frequency Time Series

Authors: Alex Tank, Emily B. Fox, Ali Shojaie

Abstract: Causal inference in multivariate time series is challenging due to the fact that the sampling rate may not be as fast as the timescale of the causal interactions. In this context, we can view our observed series as a subsampled version of the desired series. Furthermore, due to technological and other limitations, series may be observed at different sampling rates, representing a mixed frequency s… ▽ More Causal inference in multivariate time series is challenging due to the fact that the sampling rate may not be as fast as the timescale of the causal interactions. In this context, we can view our observed series as a subsampled version of the desired series. Furthermore, due to technological and other limitations, series may be observed at different sampling rates, representing a mixed frequency setting. To determine instantaneous and lagged effects between time series at the true causal scale, we take a model-based approach based on structural vector autoregressive (SVAR) models. In this context, we present a unifying framework for parameter identifiability and estimation under both subsampling and mixed frequencies when the noise, or shocks, are non-Gaussian. Importantly, by studying the SVAR case, we are able to both provide identifiability and estimation methods for the causal structure of both lagged and instantaneous effects at the desired time scale. We further derive an exact EM algorithm for inference in both subsampled and mixed frequency settings. We validate our approach in simulated scenarios and on two real world data sets. △ Less

Submitted 8 April, 2017; originally announced April 2017.

Zero-field Edge Magnetoplasmons in a Magnetic Topological Insulator

Authors: A. C. Mahoney, J. I. Colless, L. Peeters, S. J. Pauka, E. J. Fox, X. Kou, Lei Pan, K. L. Wang, D. Goldhaber-Gordon, D. J. Reilly

Abstract: Incorporating ferromagnetic dopants, such as chromium or vanadium, into thin films of the three-dimensional (3D) topological insulator (TI) (Bi,Sb)2Te3 has recently led to the realisation of the quantum anomalous Hall effect (QAHE), a unique phase of quantum matter. These materials are of great interest, since they may support electrical currents that flow without resistance via edge channels, eve… ▽ More Incorporating ferromagnetic dopants, such as chromium or vanadium, into thin films of the three-dimensional (3D) topological insulator (TI) (Bi,Sb)2Te3 has recently led to the realisation of the quantum anomalous Hall effect (QAHE), a unique phase of quantum matter. These materials are of great interest, since they may support electrical currents that flow without resistance via edge channels, even at zero magnetic field. To date, the QAHE has been investigated using low-frequency transport measurements. However, transport requires contacting the sample and results can be difficult to interpret due to the presence of parallel conductive paths, via either the bulk or surface, or because additional non-chiral edge channels may exist. Here, we move beyond transport measurements by probing the microwave response of a magnetised disk of Cr-(Bi,Sb)2Te3. We identify features associated with chiral edge magnetoplasmons (EMPs), a signature that robust edge-channels are indeed intrinsic to this material system. Our results provide a measure of the velocity of edge excitations without contacting the sample, and pave the way for a new, on-chip circuit element of practical importance: the TI, zero-field microwave circulator. △ Less

Submitted 8 March, 2017; originally announced March 2017.

Interplay of chiral and helical states in a Quantum Spin Hall Insulator lateral junction

Authors: M. R. Calvo, F. de Juan, R. Ilan, E. J. Fox, A. J. Bestwick, M. Mühlbauer, J. Wang, C. Ames, P. Leubner, C. Brüne, S. C. Zhang, H. Buhmann, L. W. Molenkamp, D. Goldhaber-Gordon

Abstract: We study the electronic transport across an electrostatically-gated lateral junction in a HgTe quantum well, a canonical 2D topological insulator, with and without applied magnetic field. We control carrier density inside and outside a junction region independently and hence tune the number and nature of 1D edge modes propagating in each of those regions. Outside the 2D gap, magnetic field drives… ▽ More We study the electronic transport across an electrostatically-gated lateral junction in a HgTe quantum well, a canonical 2D topological insulator, with and without applied magnetic field. We control carrier density inside and outside a junction region independently and hence tune the number and nature of 1D edge modes propagating in each of those regions. Outside the 2D gap, magnetic field drives the system to the quantum Hall regime, and chiral states propagate at the edge. In this regime, we observe fractional plateaus which reflect the equilibration between 1D chiral modes across the junction. As carrier density approaches zero in the central region and at moderate fields, we observe oscillations in resistance that we attribute to Fabry-Perot interference in the helical states, enabled by the broken time reversal symmetry. At higher fields, those oscillations disappear, in agreement with the expected absence of helical states when band inversion is lifted. △ Less

Submitted 13 December, 2017; v1 submitted 27 February, 2017; originally announced February 2017.

Comments: 5 pages, 4 figures, supp. mat

Journal ref: Phys. Rev. Lett. 119, 226401 (2017)

Adapting astronomical source detection software to help detect animals in thermal images obtained by unmanned aerial systems

Authors: S. N. Longmore, R. P. Collins, S. Pfeifer, S. E. Fox, M. Mulero-Pazmany, F. Bezombes, A. Goodwind, M. de Juan Ovelar, J. H. Knapen, S. A. Wich

Abstract: In this paper we describe an unmanned aerial system equipped with a thermal-infrared camera and software pipeline that we have developed to monitor animal populations for conservation purposes. Taking a multi-disciplinary approach to tackle this problem, we use freely available astronomical source detection software and the associated expertise of astronomers, to efficiently and reliably detect hu… ▽ More In this paper we describe an unmanned aerial system equipped with a thermal-infrared camera and software pipeline that we have developed to monitor animal populations for conservation purposes. Taking a multi-disciplinary approach to tackle this problem, we use freely available astronomical source detection software and the associated expertise of astronomers, to efficiently and reliably detect humans and animals in aerial thermal-infrared footage. Combining this astronomical detection software with existing machine learning algorithms into a single, automated, end-to-end pipeline, we test the software using aerial video footage taken in a controlled, field-like environment. We demonstrate that the pipeline works reliably and describe how it can be used to estimate the completeness of different observational datasets to objects of a given type as a function of height, observing conditions etc. -- a crucial step in converting video footage to scientifically useful information such as the spatial distribution and density of different animal species. Finally, having demonstrated the potential utility of the system, we describe the steps we are taking to adapt the system for work in the field, in particular systematic monitoring of endangered species at National Parks around the world. △ Less

Submitted 6 January, 2017; originally announced January 2017.

Comments: 16 pages, 6 figures, accepted to International Journal of Remote Sensing

Anatomy of Scholarly Information Behavior Patterns in the Wake of Academic Social Media Platforms

Authors: Hamed Alhoori, Mohammed Samaka, Richard Furuta, Edward A. Fox

Abstract: As more scholarly content is born digital or converted to a digital format, digital libraries are becoming increasingly vital to researchers seeking to leverage scholarly big data for scientific discovery. Although scholarly products are available in abundance-especially in environments created by the advent of social networking services-little is known about international scholarly information ne… ▽ More As more scholarly content is born digital or converted to a digital format, digital libraries are becoming increasingly vital to researchers seeking to leverage scholarly big data for scientific discovery. Although scholarly products are available in abundance-especially in environments created by the advent of social networking services-little is known about international scholarly information needs, information-seeking behavior, or information use. The purpose of this paper is to address these gaps via an in-depth analysis of the information needs and information-seeking behavior of researchers, both students and faculty, at two universities, one in the U.S. and the other in Qatar. Based on this analysis, the study identifies and describes new behavior patterns on the part of researchers as they engage in the information-seeking process. The analysis reveals that the use of academic social networks has notable effects on various scholarly activities. Further, this study identifies differences between students and faculty members in regard to their use of academic social networks, and it identifies differences between researchers according to discipline. Although the researchers who participated in the present study represent a range of disciplinary and cultural backgrounds, the study reports a number of similarities in terms of the researchers' scholarly activities. △ Less

Submitted 7 August, 2018; v1 submitted 22 December, 2016; originally announced December 2016.

The Atacama Cosmology Telescope: Two-Season ACTPol Lensing Power Spectrum

Authors: Blake D. Sherwin, Alexander van Engelen, Neelima Sehgal, Mathew Madhavacheril, Graeme E. Addison, Simone Aiola, Rupert Allison, Nicholas Battaglia, James A. Beall, Daniel T. Becker, J. Richard Bond, Erminia Calabrese, Rahul Datta, Mark J. Devlin, Rolando Dunner, Joanna Dunkley, Anna E. Fox, Patricio Gallardo, Mark Halpern, Matthew Hasselfield, Shawn Henderson, J. Colin Hill, Gene C. Hilton, Johannes Hubmayr, John P. Hughes , et al. (21 additional authors not shown)

Abstract: We report a measurement of the power spectrum of cosmic microwave background (CMB) lensing from two seasons of Atacama Cosmology Telescope Polarimeter (ACTPol) CMB data. The CMB lensing power spectrum is extracted from both temperature and polarization data using quadratic estimators. We obtain results that are consistent with the expectation from the best-fit Planck LCDM model over a range of mul… ▽ More We report a measurement of the power spectrum of cosmic microwave background (CMB) lensing from two seasons of Atacama Cosmology Telescope Polarimeter (ACTPol) CMB data. The CMB lensing power spectrum is extracted from both temperature and polarization data using quadratic estimators. We obtain results that are consistent with the expectation from the best-fit Planck LCDM model over a range of multipoles L=80-2100, with an amplitude of lensing A_lens = 1.06 +/- 0.15 (stat.) +/- 0.06 (sys.) relative to Planck. Our measurement of the CMB lensing power spectrum gives sigma_8 Omega_m^0.25 = 0.643 +/- 0.054; including baryon acoustic oscillation scale data, we constrain the amplitude of density fluctuations to be sigma_8 = 0.831 +/- 0.053. We also update constraints on the neutrino mass sum. We verify our lensing measurement with a number of null tests and systematic checks, finding no evidence of significant systematic errors. This measurement relies on a small fraction of the ACTPol data already taken; more precise lensing results can therefore be expected from the full ACTPol dataset. △ Less

Submitted 29 November, 2016; originally announced November 2016.

Comments: 17 pages, 11 figures, to be submitted to Physical Review D

Journal ref: Phys. Rev. D 95, 123529 (2017)

The Atacama Cosmology Telescope: Two-Season ACTPol Spectra and Parameters

Authors: Thibaut Louis, Emily Grace, Matthew Hasselfield, Marius Lungu, Loïc Maurin, Graeme E. Addison, Peter A. R. Ade, Simone Aiola, Rupert Allison, Mandana Amiri, Elio Angile, Nicholas Battaglia, James A. Beall, Francesco de Bernardis, J. Richard Bond, Joe Britton, Erminia Calabrese, Hsiao-mei Cho, Steve K. Choi, Kevin Coughlin, Devin Crichton, Kevin Crowley, Rahul Datta, Mark J. Devlin, Simon R. Dicker , et al. (58 additional authors not shown)

Abstract: We present the temperature and polarization angular power spectra measured by the Atacama Cosmology Telescope Polarimeter (ACTPol). We analyze night-time data collected during 2013-14 using two detector arrays at 149 GHz, from 548 deg$^2$ of sky on the celestial equator. We use these spectra, and the spectra measured with the MBAC camera on ACT from 2008-10, in combination with Planck and WMAP dat… ▽ More We present the temperature and polarization angular power spectra measured by the Atacama Cosmology Telescope Polarimeter (ACTPol). We analyze night-time data collected during 2013-14 using two detector arrays at 149 GHz, from 548 deg$^2$ of sky on the celestial equator. We use these spectra, and the spectra measured with the MBAC camera on ACT from 2008-10, in combination with Planck and WMAP data to estimate cosmological parameters from the temperature, polarization, and temperature-polarization cross-correlations. We find the new ACTPol data to be consistent with the LCDM model. The ACTPol temperature-polarization cross-spectrum now provides stronger constraints on multiple parameters than the ACTPol temperature spectrum, including the baryon density, the acoustic peak angular scale, and the derived Hubble constant. Adding the new data to planck temperature data tightens the limits on damping tail parameters, for example reducing the joint uncertainty on the number of neutrino species and the primordial helium fraction by 20%. △ Less

Submitted 7 October, 2016; originally announced October 2016.

Comments: 23 pages, 25 figures

Irreversible Samplers from Jump and Continuous Markov Processes

Authors: Yi-An Ma, Emily B. Fox, Tianqi Chen, Lei Wu

Abstract: In this paper, we propose irreversible versions of the Metropolis Hastings (MH) and Metropolis adjusted Langevin algorithm (MALA) with a main focus on the latter. For the former, we show how one can simply switch between different proposal and acceptance distributions upon rejection to obtain an irreversible jump sampler (I-Jump). The resulting algorithm has a simple implementation akin to MH, but… ▽ More In this paper, we propose irreversible versions of the Metropolis Hastings (MH) and Metropolis adjusted Langevin algorithm (MALA) with a main focus on the latter. For the former, we show how one can simply switch between different proposal and acceptance distributions upon rejection to obtain an irreversible jump sampler (I-Jump). The resulting algorithm has a simple implementation akin to MH, but with the demonstrated benefits of irreversibility. We then show how the previously proposed MALA method can also be extended to exploit irreversible stochastic dynamics as proposal distributions in the I-Jump sampler. Our experiments explore how irreversibility can increase the efficiency of the samplers in different situations. △ Less

Submitted 12 March, 2018; v1 submitted 21 August, 2016; originally announced August 2016.

Journal ref: Stat. Comput. (2018) 1-26

The Atacama Cosmology Telescope: The polarization-sensitive ACTPol instrument

Authors: R. J. Thornton, P. A. R. Ade, S. Aiola, F. E. Angile, M. Amiri, J. A. Beall, D. T. Becker, H-M. Cho, S. K. Choi, P. Corlies, K. P. Coughlin, R. Datta, M. J. Devlin, S. R. Dicker, R. Dunner, J. W. Fowler, A. E. Fox, P. A. Gallardo, J. Gao, E. Grace, M. Halpern, M. Hasselfield, S. W. Henderson, G. C. Hilton, A. D. Hincks , et al. (31 additional authors not shown)

Abstract: The Atacama Cosmology Telescope (ACT) is designed to make high angular resolution measurements of anisotropies in the Cosmic Microwave Background (CMB) at millimeter wavelengths. We describe ACTPol, an upgraded receiver for ACT, which uses feedhorn-coupled, polarization-sensitive detector arrays, a 3 degree field of view, 100 mK cryogenics with continuous cooling, and meta material anti-reflection… ▽ More The Atacama Cosmology Telescope (ACT) is designed to make high angular resolution measurements of anisotropies in the Cosmic Microwave Background (CMB) at millimeter wavelengths. We describe ACTPol, an upgraded receiver for ACT, which uses feedhorn-coupled, polarization-sensitive detector arrays, a 3 degree field of view, 100 mK cryogenics with continuous cooling, and meta material anti-reflection coatings. ACTPol comprises three arrays with separate cryogenic optics: two arrays at a central frequency of 148 GHz and one array operating simultaneously at both 97 GHz and 148 GHz. The combined instrument sensitivity, angular resolution, and sky coverage are optimized for measuring angular power spectra, clusters via the thermal Sunyaev-Zel'dovich and kinetic Sunyaev-Zel'dovich signals, and CMB lensing due to large scale structure. The receiver was commissioned with its first 148 GHz array in 2013, observed with both 148 GHz arrays in 2014, and has recently completed its first full season of operations with the full suite of three arrays. This paper provides an overview of the design and initial performance of the receiver and related systems. △ Less

Submitted 20 May, 2016; originally announced May 2016.

A Complete Recipe for Stochastic Gradient MCMC

Authors: Yi-An Ma, Tianqi Chen, Emily B. Fox

Abstract: Many recent Markov chain Monte Carlo (MCMC) samplers leverage continuous dynamics to define a transition kernel that efficiently explores a target distribution. In tandem, a focus has been on devising scalable variants that subsample the data and use stochastic gradients in place of full-data gradients in the dynamic simulations. However, such stochastic gradient MCMC samplers have lagged behind t… ▽ More Many recent Markov chain Monte Carlo (MCMC) samplers leverage continuous dynamics to define a transition kernel that efficiently explores a target distribution. In tandem, a focus has been on devising scalable variants that subsample the data and use stochastic gradients in place of full-data gradients in the dynamic simulations. However, such stochastic gradient MCMC samplers have lagged behind their full-data counterparts in terms of the complexity of dynamics considered since proving convergence in the presence of the stochastic gradient noise is non-trivial. Even with simple dynamics, significant physical intuition is often required to modify the dynamical system to account for the stochastic gradient noise. In this paper, we provide a general recipe for constructing MCMC samplers--including stochastic gradient versions--based on continuous Markov processes specified via two matrices. We constructively prove that the framework is complete. That is, any continuous Markov process that provides samples from the target distribution can be written in our framework. We show how previous continuous-dynamic samplers can be trivially "reinvented" in our framework, avoiding the complicated sampler-specific proofs. We likewise use our recipe to straightforwardly propose a new state-adaptive sampler: stochastic gradient Riemann Hamiltonian Monte Carlo (SGRHMC). Our experiments on simulated data and a streaming Wikipedia analysis demonstrate that the proposed SGRHMC sampler inherits the benefits of Riemann HMC, with the scalability of stochastic gradient methods. △ Less

Submitted 31 October, 2015; v1 submitted 15 June, 2015; originally announced June 2015.

Bayesian Structure Learning for Stationary Time Series

Authors: Alex Tank, Nicholas Foti, Emily Fox

Abstract: While much work has explored probabilistic graphical models for independent data, less attention has been paid to time series. The goal in this setting is to determine conditional independence relations between entire time series, which for stationary series, are encoded by zeros in the inverse spectral density matrix. We take a Bayesian approach to structure learning, placing priors on (i) the gr… ▽ More While much work has explored probabilistic graphical models for independent data, less attention has been paid to time series. The goal in this setting is to determine conditional independence relations between entire time series, which for stationary series, are encoded by zeros in the inverse spectral density matrix. We take a Bayesian approach to structure learning, placing priors on (i) the graph structure and (ii) spectral matrices given the graph. We leverage a Whittle likelihood approximation and define a conjugate prior---the hyper complex inverse Wishart---on the complex-valued and graph-constrained spectral matrices. Due to conjugacy, we can analytically marginalize the spectral matrices and obtain a closed-form marginal likelihood of the time series given a graph. Importantly, our analytic marginal likelihood allows us to avoid inference of the complex spectral matrices themselves and places us back into the framework of standard (Bayesian) structure learning. In particular, combining this marginal likelihood with our graph prior leads to efficient inference of the time series graph itself, which we base on a stochastic search procedure, though any standard approach can be straightforwardly modified to our time series case. We demonstrate our methods on analyzing stock data and neuroimaging data of brain activity during various auditory tasks. △ Less

Submitted 3 July, 2015; v1 submitted 12 May, 2015; originally announced May 2015.

Achieving a Hyperlocal Housing Price Index: Overcoming Data Sparsity by Bayesian Dynamical Modeling of Multiple Data Streams

Authors: You Ren, Emily B. Fox, Andrew Bruce

Abstract: Understanding how housing values evolve over time is important to policy makers, consumers and real estate professionals. Existing methods for constructing housing indices are computed at a coarse spatial granularity, such as metropolitan regions, which can mask or distort price dynamics apparent in local markets, such as neighborhoods and census tracts. A challenge in moving to estimates at, for… ▽ More Understanding how housing values evolve over time is important to policy makers, consumers and real estate professionals. Existing methods for constructing housing indices are computed at a coarse spatial granularity, such as metropolitan regions, which can mask or distort price dynamics apparent in local markets, such as neighborhoods and census tracts. A challenge in moving to estimates at, for example, the census tract level is the sparsity of spatiotemporally localized house sales observations. Our work aims at addressing this challenge by leveraging observations from multiple census tracts discovered to have correlated valuation dynamics. Our proposed Bayesian nonparametric approach builds on the framework of latent factor models to enable a flexible, data-driven method for inferring the clustering of correlated census tracts. We explore methods for scalability and parallelizability of computations, yielding a housing valuation index at the level of census tract rather than zip code, and on a monthly basis rather than quarterly. Our analysis is provided on a large Seattle metropolitan housing dataset. △ Less

Submitted 5 May, 2015; originally announced May 2015.

Resonant magneto-optic Kerr effect in the magnetic topological insulator Cr:(Sb$_x$,Bi$_{1-x}$)$_2$Te$_3$

Authors: Shreyas Patankar, J. P. Hinton, Joel Griesmar, J. Orenstein, J. S. Dodge, Xufeng Kou, Lei Pan, Kang L. Wang, A. J. Bestwick, E. J. Fox, D. Goldhaber-Gordon, Jing Wang, Shou-Cheng Zhang

Abstract: We report measurements of the polar Kerr effect, proportional to the out-of-plane component of the magnetization, in thin films of the magnetically doped topological insulator $(\text{Cr}_{0.12}\text{Bi}_{0.26}\text{Sb}_{0.62})_2\text{Te}_3$. Measurements of the complex Kerr angle, $Θ_K$, were performed as a function of photon energy in the range $0.8\text{ eV}<\hbarω<3.0\text{ eV}$. We observed a… ▽ More We report measurements of the polar Kerr effect, proportional to the out-of-plane component of the magnetization, in thin films of the magnetically doped topological insulator $(\text{Cr}_{0.12}\text{Bi}_{0.26}\text{Sb}_{0.62})_2\text{Te}_3$. Measurements of the complex Kerr angle, $Θ_K$, were performed as a function of photon energy in the range $0.8\text{ eV}<\hbarω<3.0\text{ eV}$. We observed a peak in the real part of $Θ_K(ω)$ and zero crossing in the imaginary part that we attribute to resonant interaction with a spin-orbit avoided crossing located $\approx$ 1.6 eV above the Fermi energy. The resonant enhancement allows measurement of the temperature and magnetic field dependence of $Θ_K$ in the ultrathin film limit, $d\geq2$ quintuple layers. We find a sharp transition to zero remanent magnetization at 6 K for $d<8$~QL, consistent with theories of the dependence of impurity spin interactions on film thickness and their location relative to topological insulator surfaces. △ Less

Submitted 2 December, 2015; v1 submitted 4 May, 2015; originally announced May 2015.

Comments: 6 pages, 5 figures

Journal ref: Phys. Rev. B 92, 214440 (2015)

Precise quantization of anomalous Hall effect near zero magnetic field

Authors: A. J. Bestwick, E. J. Fox, Xufeng Kou, Lei Pan, Kang L. Wang, D. Goldhaber-Gordon

Abstract: We report a nearly ideal quantum anomalous Hall effect in a three-dimensional topological insulator thin film with ferromagnetic doping. Near zero applied magnetic field we measure exact quantization in Hall resistance to within a part per 10,000 and longitudinal resistivity under 1 ohm per square, with chiral edge transport explicitly confirmed by non-local measurements. Deviations from this beha… ▽ More We report a nearly ideal quantum anomalous Hall effect in a three-dimensional topological insulator thin film with ferromagnetic doping. Near zero applied magnetic field we measure exact quantization in Hall resistance to within a part per 10,000 and longitudinal resistivity under 1 ohm per square, with chiral edge transport explicitly confirmed by non-local measurements. Deviations from this behavior are found to be caused by thermally-activated carriers, which can be eliminated by taking advantage of an unexpected magnetocaloric effect. △ Less

Submitted 1 April, 2015; v1 submitted 9 December, 2014; originally announced December 2014.

Comments: 5 pages, 4 figures, with 9 pages of supplementary information

Journal ref: Phys. Rev. Lett. 114, 187201 (2015)

Streaming Variational Inference for Bayesian Nonparametric Mixture Models

Authors: Alex Tank, Nicholas J. Foti, Emily B. Fox

Abstract: In theory, Bayesian nonparametric (BNP) models are well suited to streaming data scenarios due to their ability to adapt model complexity with the observed data. Unfortunately, such benefits have not been fully realized in practice; existing inference algorithms are either not applicable to streaming applications or not extensible to BNP models. For the special case of Dirichlet processes, streami… ▽ More In theory, Bayesian nonparametric (BNP) models are well suited to streaming data scenarios due to their ability to adapt model complexity with the observed data. Unfortunately, such benefits have not been fully realized in practice; existing inference algorithms are either not applicable to streaming applications or not extensible to BNP models. For the special case of Dirichlet processes, streaming inference has been considered. However, there is growing interest in more flexible BNP models building on the class of normalized random measures (NRMs). We work within this general framework and present a streaming variational inference algorithm for NRM mixture models. Our algorithm is based on assumed density filtering (ADF), leading straightforwardly to expectation propagation (EP) for large-scale batch inference as well. We demonstrate the efficacy of the algorithm on clustering documents in large, streaming text corpora. △ Less

Submitted 21 April, 2015; v1 submitted 1 December, 2014; originally announced December 2014.

Stochastic Variational Inference for Hidden Markov Models

Authors: Nicholas J. Foti, Jason Xu, Dillon Laird, Emily B. Fox

Abstract: Variational inference algorithms have proven successful for Bayesian analysis in large data settings, with recent advances using stochastic variational inference (SVI). However, such methods have largely been studied in independent or exchangeable data settings. We develop an SVI algorithm to learn the parameters of hidden Markov models (HMMs) in a time-dependent data setting. The challenge in app… ▽ More Variational inference algorithms have proven successful for Bayesian analysis in large data settings, with recent advances using stochastic variational inference (SVI). However, such methods have largely been studied in independent or exchangeable data settings. We develop an SVI algorithm to learn the parameters of hidden Markov models (HMMs) in a time-dependent data setting. The challenge in applying stochastic optimization in this setting arises from dependencies in the chain, which must be broken to consider minibatches of observations. We propose an algorithm that harnesses the memory decay of the chain to adaptively bound errors arising from edge effects. We demonstrate the effectiveness of our algorithm on synthetic experiments and a large genomics dataset where a batch algorithm is computationally infeasible. △ Less

Submitted 6 November, 2014; originally announced November 2014.

Comments: Appears in Advances in Neural Information Processing Systems (NIPS), 2014