-
Causal Climate Emulation with Bayesian Filtering
Authors:
Sebastian Hickman,
Ilija Trajkovic,
Julia Kaltenborn,
Francis Pelletier,
Alex Archibald,
Yaniv Gurwicz,
Peer Nowack,
David Rolnick,
Julien Boussard
Abstract:
Traditional models of climate change use complex systems of coupled equations to simulate physical processes across the Earth system. These simulations are highly computationally expensive, limiting our predictions of climate change and analyses of its causes and effects. Machine learning has the potential to quickly emulate data from climate models, but current approaches are not able to incorpor…
▽ More
Traditional models of climate change use complex systems of coupled equations to simulate physical processes across the Earth system. These simulations are highly computationally expensive, limiting our predictions of climate change and analyses of its causes and effects. Machine learning has the potential to quickly emulate data from climate models, but current approaches are not able to incorporate physics-informed causal relationships. Here, we develop an interpretable climate model emulator based on causal representation learning. We derive a physics-informed approach including a Bayesian filter for stable long-term autoregressive emulation. We demonstrate that our emulator learns accurate climate dynamics, and we show the importance of each one of its components on a realistic synthetic dataset and data from two widely deployed climate models.
△ Less
Submitted 11 June, 2025;
originally announced June 2025.
-
Flat panel laser displays enabled by large-scale visible photonic integrated circuits
Authors:
Zhujun Shi,
Risheng Cheng,
Guohua Wei,
Steven A. Hickman,
Min Chul Shin,
Peter Topalian,
Lei Wang,
Dusan Coso,
Brian Le,
Lizzy Lee,
Sean Braxton,
Alexander Koshelev,
Maxwell F. Parsons,
Rahul Agarwal,
Barry Silverstein,
Yun Wang,
Giuseppe Calafiore
Abstract:
Laser-based displays are highly sought after for their superior brightness and color performance, especially in advanced applications like augmented reality (AR). However, their broader adoption has been hindered by bulky projector designs and complex optical module assemblies. Here, we introduce a new laser display architecture enabled by large-scale visible photonic integrated circuits (PICs) to…
▽ More
Laser-based displays are highly sought after for their superior brightness and color performance, especially in advanced applications like augmented reality (AR). However, their broader adoption has been hindered by bulky projector designs and complex optical module assemblies. Here, we introduce a new laser display architecture enabled by large-scale visible photonic integrated circuits (PICs) to address these challenges. Unlike previous projector-style laser displays, this architecture features an ultra-thin, flat-panel form factor, replacing bulky free-space illumination modules with a single, high-performance photonic chip. Centimeter-scale PIC devices, which integrate thousands of distinct optical components on-chip, are carefully tailored to achieve high display uniformity, contrast, and efficiency. We demonstrate a 2 mm-thick flat-panel laser display combining the PIC with a liquid-crystal-on-silicon (LCoS) panel, achieving 211% of the color gamut and more than 80% volume reduction compared to traditional LCoS displays. We further showcase its application in a see-through AR system. Our work represents a major advancement in the integration of nanophotonics with display technology, enabling a range of new display concepts, from high-performance immersive displays to slim-panel 3D holography.
△ Less
Submitted 26 December, 2024;
originally announced December 2024.
-
MammoDG: Generalisable Deep Learning Breaks the Limits of Cross-Domain Multi-Center Breast Cancer Screening
Authors:
Yijun Yang,
Shujun Wang,
Lihao Liu,
Sarah Hickman,
Fiona J Gilbert,
Carola-Bibiane Schönlieb,
Angelica I. Aviles-Rivero
Abstract:
Breast cancer is a major cause of cancer death among women, emphasising the importance of early detection for improved treatment outcomes and quality of life. Mammography, the primary diagnostic imaging test, poses challenges due to the high variability and patterns in mammograms. Double reading of mammograms is recommended in many screening programs to improve diagnostic accuracy but increases ra…
▽ More
Breast cancer is a major cause of cancer death among women, emphasising the importance of early detection for improved treatment outcomes and quality of life. Mammography, the primary diagnostic imaging test, poses challenges due to the high variability and patterns in mammograms. Double reading of mammograms is recommended in many screening programs to improve diagnostic accuracy but increases radiologists' workload. Researchers explore Machine Learning models to support expert decision-making. Stand-alone models have shown comparable or superior performance to radiologists, but some studies note decreased sensitivity with multiple datasets, indicating the need for high generalisation and robustness models. This work devises MammoDG, a novel deep-learning framework for generalisable and reliable analysis of cross-domain multi-center mammography data. MammoDG leverages multi-view mammograms and a novel contrastive mechanism to enhance generalisation capabilities. Extensive validation demonstrates MammoDG's superiority, highlighting the critical importance of domain generalisation for trustworthy mammography analysis in imaging protocol variations.
△ Less
Submitted 2 August, 2023;
originally announced August 2023.
-
Precise Blaze Angle Measurements of Lithographically Fabricated Silicon Immersion Gratings
Authors:
Emily Lubar,
Daniel T. Jaffe,
Cynthia Brooks,
Sierra Hickman,
Michael Gully-Santiago,
Gregory Mace
Abstract:
Silicon immersion gratings and grisms enable compact, near-infrared spectrographs with high throughput. These instruments find use in ground-based efforts to characterize stellar and exoplanet atmospheres, and in space-based observatories. Our grating fabrication technique uses x-ray crystallography to orient silicon parts prior to cutting, followed by lithography and wet chemical etching to produ…
▽ More
Silicon immersion gratings and grisms enable compact, near-infrared spectrographs with high throughput. These instruments find use in ground-based efforts to characterize stellar and exoplanet atmospheres, and in space-based observatories. Our grating fabrication technique uses x-ray crystallography to orient silicon parts prior to cutting, followed by lithography and wet chemical etching to produce the blaze. This process takes advantage of the crystal structure and relative difference in etching rates between the (100) and (111) planes such that we can produce parts that have surface errors < λ/4. Previous measurements indicate that chemical etching can yield a final etched blaze that slightly differs from the orientation of the (111) plane. This difference can be corrected by the mechanical mount in the case of the immersion gratings, but doing so may compromise grating throughput due to shadowing. In the case of the grisms, failure to take the actual blaze into account will reduce the brightness of the undeviated ray. We report on multiple techniques to precisely measure the blaze of our in-house fabricated immersion gratings. The first method uses a scanning electron microscope to image the blaze profile, which yields a measurement precision of 0.5 degrees. The second method is an optical method of measuring the angle between blaze faces using a rotation stage, which yields a measurement precision of 0.2 degrees. Finally, we describe a theoretical blaze function modeling method, which we expect to yield a measurement precision of 0.1 degrees. With these methods, we can quantify the accuracy with which the wet etching produces the required blaze and further optimize grating and grism efficiencies.
△ Less
Submitted 14 December, 2020;
originally announced December 2020.
-
Direct Methods and Symbolic Software for Conservation Laws of Nonlinear Equations
Authors:
Willy Hereman,
Paul J. Adams,
Holly L. Eklund,
Mark S. Hickman,
Barend M. Herbst
Abstract:
We present direct methods and symbolic software for the computation of conservation laws of nonlinear partial differential equations (PDEs) and differential-difference equations (DDEs).The methods are applied to nonlinear PDEs in (1+1) dimensions with polynomial nonlinearities which include the Korteweg-de Vries (KdV), Boussinesq, and Drinfel'd-Sokolov-Wilson equations. An adaptation of the meth…
▽ More
We present direct methods and symbolic software for the computation of conservation laws of nonlinear partial differential equations (PDEs) and differential-difference equations (DDEs).The methods are applied to nonlinear PDEs in (1+1) dimensions with polynomial nonlinearities which include the Korteweg-de Vries (KdV), Boussinesq, and Drinfel'd-Sokolov-Wilson equations. An adaptation of the methods is applied to PDEs with transcendental nonlinearities. Examples include the sine-Gordon, sinh-Gordon, and Liouville equations.
With respect to nonlinear DDEs, our methods are applied to Kac-van Moerbeke, Toda, and Ablowitz-Ladik lattices. To overcome the shortcomings of the undetermined coefficients method, we designed a new direct method which uses leading order analysis. That method is applied to discretizations of the KdV and modified KdV equations, and a combination thereof. Additional examples include lattices due to Bogoyavlenskii, Belov-Chaltikian, and Blaszak-Marciniak.
The undetermined coefficient methods for PDEs and DDEs have been implemented in Mathematica. The code "TransPDEDensityFlux.m" computes densities and fluxes of systems of PDEs with or without transcendental nonlinearities. The code "DDEDensityFlux.m" does the same for polynomial nonlinear DDEs. Starting from the leading order terms, the new Maple library "discrete" computes densities and fluxes of nonlinear DDEs.
△ Less
Submitted 1 March, 2008;
originally announced March 2008.
-
Continuous and Discrete Homotopy Operators with Applications in Integrability Testing
Authors:
Willy Hereman,
Michael Colagrosso,
Ryan Sayers,
Adam Ringler,
Bernard Deconinck,
Michael Nivala,
Mark S. Hickman
Abstract:
We introduce calculus-based formulas for the continuous Euler and homotopy operators. The 1D continuous homotopy operator automates integration by parts on the jet space. Its 3D generalization allows one to invert the total divergence operator. As a practical application, we show how the operators can be used to symbolically compute local conservation laws of nonlinear systems of partial differe…
▽ More
We introduce calculus-based formulas for the continuous Euler and homotopy operators. The 1D continuous homotopy operator automates integration by parts on the jet space. Its 3D generalization allows one to invert the total divergence operator. As a practical application, we show how the operators can be used to symbolically compute local conservation laws of nonlinear systems of partial differential equations in multi-dimensions.
By analogy to the continuous case, we also present concrete formulas for the discrete Euler and homotopy operators. Essentially, the discrete homotopy operator carries out summation by parts. We use it to algorithmically invert the forward difference operator. We apply the discrete operator to compute fluxes of differential-difference equations in (1+1) dimensions.
Our calculus-based approach allows for a straightforward implementation of the operators in major computer algebra system, such as Mathematica and Maple. The symbolic algorithms for integration and summation by parts are illustrated with elementary examples. The algorithms to compute conservation laws are illustrated with nonlinear PDEs and their discretizations arising in fluid dynamics and mathematical physics.
△ Less
Submitted 19 January, 2005;
originally announced January 2005.
