-
Simplifying Bayesian Optimization Via In-Context Direct Optimum Sampling
Authors:
Gustavo Sutter Pessurno de Carvalho,
Mohammed Abdulrahman,
Hao Wang,
Sriram Ganapathi Subramanian,
Marc St-Aubin,
Sharon O'Sullivan,
Lawrence Wan,
Luis Ricardez-Sandoval,
Pascal Poupart,
Agustinus Kristiadi
Abstract:
The optimization of expensive black-box functions is ubiquitous in science and engineering. A common solution to this problem is Bayesian optimization (BO), which is generally comprised of two components: (i) a surrogate model and (ii) an acquisition function, which generally require expensive re-training and optimization steps at each iteration, respectively. Although recent work enabled in-conte…
▽ More
The optimization of expensive black-box functions is ubiquitous in science and engineering. A common solution to this problem is Bayesian optimization (BO), which is generally comprised of two components: (i) a surrogate model and (ii) an acquisition function, which generally require expensive re-training and optimization steps at each iteration, respectively. Although recent work enabled in-context surrogate models that do not require re-training, virtually all existing BO methods still require acquisition function maximization to select the next observation, which introduces many knobs to tune, such as Monte Carlo samplers and multi-start optimizers. In this work, we propose a completely in-context, zero-shot solution for BO that does not require surrogate fitting or acquisition function optimization. This is done by using a pre-trained deep generative model to directly sample from the posterior over the optimum point. We show that this process is equivalent to Thompson sampling and demonstrate the capabilities and cost-effectiveness of our foundation model on a suite of real-world benchmarks. We achieve an efficiency gain of more than 35x in terms of wall-clock time when compared with Gaussian process-based BO, enabling efficient parallel and distributed BO, e.g., for high-throughput optimization.
△ Less
Submitted 29 May, 2025;
originally announced May 2025.
-
Learning multiplane images from single views with self-supervision
Authors:
Gustavo Sutter P. Carvalho,
Diogo C. Luvizon,
Antonio Joia Neto,
Andre G. C. Pacheco,
Otavio A. B. Penatti
Abstract:
Generating static novel views from an already captured image is a hard task in computer vision and graphics, in particular when the single input image has dynamic parts such as persons or moving objects. In this paper, we tackle this problem by proposing a new framework, called CycleMPI, that is capable of learning a multiplane image representation from single images through a cyclic training stra…
▽ More
Generating static novel views from an already captured image is a hard task in computer vision and graphics, in particular when the single input image has dynamic parts such as persons or moving objects. In this paper, we tackle this problem by proposing a new framework, called CycleMPI, that is capable of learning a multiplane image representation from single images through a cyclic training strategy for self-supervision. Our framework does not require stereo data for training, therefore it can be trained with massive visual data from the Internet, resulting in a better generalization capability even for very challenging cases. Although our method does not require stereo data for supervision, it reaches results on stereo datasets comparable to the state of the art in a zero-shot scenario. We evaluated our method on RealEstate10K and Mannequin Challenge datasets for view synthesis and presented qualitative results on Places II dataset.
△ Less
Submitted 19 October, 2021; v1 submitted 18 October, 2021;
originally announced October 2021.
-
Adaptive Multiplane Image Generation from a Single Internet Picture
Authors:
Diogo C. Luvizon,
Gustavo Sutter P. Carvalho,
Andreza A. dos Santos,
Jhonatas S. Conceicao,
Jose L. Flores-Campana,
Luis G. L. Decker,
Marcos R. Souza,
Helio Pedrini,
Antonio Joia,
Otavio A. B. Penatti
Abstract:
In the last few years, several works have tackled the problem of novel view synthesis from stereo images or even from a single picture. However, previous methods are computationally expensive, specially for high-resolution images. In this paper, we address the problem of generating a multiplane image (MPI) from a single high-resolution picture. We present the adaptive-MPI representation, which all…
▽ More
In the last few years, several works have tackled the problem of novel view synthesis from stereo images or even from a single picture. However, previous methods are computationally expensive, specially for high-resolution images. In this paper, we address the problem of generating a multiplane image (MPI) from a single high-resolution picture. We present the adaptive-MPI representation, which allows rendering novel views with low computational requirements. To this end, we propose an adaptive slicing algorithm that produces an MPI with a variable number of image planes. We present a new lightweight CNN for depth estimation, which is learned by knowledge distillation from a larger network. Occluded regions in the adaptive-MPI are inpainted also by a lightweight CNN. We show that our method is capable of producing high-quality predictions with one order of magnitude less parameters compared to previous approaches. The robustness of our method is evidenced on challenging pictures from the Internet.
△ Less
Submitted 26 November, 2020;
originally announced November 2020.
-
Chemical abundances of Seyfert 2 AGNs $-$II. N2 metallicity calibration based on SDSS
Authors:
S. P. Carvalho,
O. L. Dors,
M. V. Cardaci,
G. F. Hagele,
A. C. Krabbe,
E. Perez-Montero,
A. F. Monteiro,
M. Armah,
P. Freitas-Lemes
Abstract:
We present a semi-empirical calibration between the metallicity ($Z$) of Seyfert 2 Active Galactic Nuclei and the $N2$=log([N II]$λ$6584/H$α$) emission-line intensity ratio. This calibration was derived through the [O III]$λ$5007/[O II]$λ$3727 versus $N2$ diagram containing observational data and photoionization model results obtained with the Cloudy code. The observational sample consists of 463…
▽ More
We present a semi-empirical calibration between the metallicity ($Z$) of Seyfert 2 Active Galactic Nuclei and the $N2$=log([N II]$λ$6584/H$α$) emission-line intensity ratio. This calibration was derived through the [O III]$λ$5007/[O II]$λ$3727 versus $N2$ diagram containing observational data and photoionization model results obtained with the Cloudy code. The observational sample consists of 463 confirmed Seyfert 2 nuclei (redshift $z < 0.4$) taken from the Sloan Digital Sky Survey DR7 dataset. The obtained $Z$-$N2$ relation is valid for the range $0.3 < (Z/Z_{\odot}) < 2.0$ which corresponds to $-0.7 < \: (N2) < 0.6$. The effects of varying the ionization parameter ($U$), electron density and the slope of the spectral energy distribution on the $Z$ estimations are of the order of the uncertainty produced by the error measurements of $N2$. This result indicates the large reliability of our $Z-N2$ calibration. A relation between $U$ and the [O III]/[O II] line ratio, almost independent of other nebular parameter, was obtained.
△ Less
Submitted 20 January, 2020;
originally announced January 2020.
-
The First Birkhoff Coefficient and the Stability of 2-Periodic Orbits on Billiards
Authors:
Sylvie Oliffson Kamphorst,
Sonia Pinto de Carvalho
Abstract:
In this work we address the question of proving the stability of elliptic 2-periodic orbits for strictly convex billiards. Eventhough it is part of a widely accepted belief that ellipticity implies stability, classical theorems show that the certainty of stability relies upon more fine conditions. We present a review of the main results and general theorems and describe the procedure to fullfill…
▽ More
In this work we address the question of proving the stability of elliptic 2-periodic orbits for strictly convex billiards. Eventhough it is part of a widely accepted belief that ellipticity implies stability, classical theorems show that the certainty of stability relies upon more fine conditions. We present a review of the main results and general theorems and describe the procedure to fullfill the supplementary conditions for strictly convex billiards.
△ Less
Submitted 14 October, 2004;
originally announced October 2004.
-
Elliptic Islands on Strictly Convex Billiards
Authors:
Mario Jorge Dias Carneiro,
Sylvie Oliffson Kamphorst,
Sonia Pinto De Carvalho
Abstract:
This paper addresses the question of genericity of existence of elliptic islands for the billiard map associated to strictly convex closed curves. More precisely, we study 2-periodic orbits of billiards associated to C5 closed and strictly convex curves and show that the existence of elliptic islands is a dense property on the subset of those billiards having an elliptic 2-periodic point. Our ma…
▽ More
This paper addresses the question of genericity of existence of elliptic islands for the billiard map associated to strictly convex closed curves. More precisely, we study 2-periodic orbits of billiards associated to C5 closed and strictly convex curves and show that the existence of elliptic islands is a dense property on the subset of those billiards having an elliptic 2-periodic point. Our main tools are normal perturbations, the Birkhoff Normal Form for elliptic fixed points and Moser's Twist Theorem.
△ Less
Submitted 11 January, 2002;
originally announced January 2002.
-
Elliptic Islands on the Elliptical Stadium
Authors:
Sylvie Oliffson Kamphorst,
Sonia Pinto de Carvalho
Abstract:
We investigate the existence of elliptic islands for a special family of periodic orbits of a two-parameter family of maps corresponding to the billiard problem on the elliptical stadium. The hyperbolic or elliptical character of these orbits is also investigated. Depending on the parameters, we obtain upper bounds of ellipticity for this special family as a lower bound for chaos. On a different…
▽ More
We investigate the existence of elliptic islands for a special family of periodic orbits of a two-parameter family of maps corresponding to the billiard problem on the elliptical stadium. The hyperbolic or elliptical character of these orbits is also investigated. Depending on the parameters, we obtain upper bounds of ellipticity for this special family as a lower bound for chaos. On a different region of the parameter space, we can prove that there is no upper bound for the existence of elliptic islands. The main results we use are Birkhoff Normal Form and Moser's Twist Theorem.
△ Less
Submitted 22 September, 2000;
originally announced September 2000.
-
Bounded Gain of Energy on the Breathing Circle Billiard
Authors:
Sylvie Oliffson Kamphorst,
Sonia Pinto de Carvalho
Abstract:
The Breathing Circle is a 2-dimensional generalization of the Fermi Accelerator. It is shown that the billiard map associated to this model has invariant curves in phase space, implying that any particle will have bounded gain of energy.
The Breathing Circle is a 2-dimensional generalization of the Fermi Accelerator. It is shown that the billiard map associated to this model has invariant curves in phase space, implying that any particle will have bounded gain of energy.
△ Less
Submitted 23 November, 1998;
originally announced November 1998.
-
A Lower Bound for Chaos on the Elliptical Stadium
Authors:
Eduardo Canale,
Roberto Markarian,
Sylvie Oliffson Kamphorst,
Sonia Pinto de Carvalho
Abstract:
The elliptical stadium is a plane region bounded by a curve constructed by joining two half-ellipses by two parallel segments of equal length. The billiard inside it, as a map, generates a two parameters family of dynamical systems. It is known that the system is ergodic for a certain region of the parameter space. In this work we study the stability of a particular family of periodic orbits obt…
▽ More
The elliptical stadium is a plane region bounded by a curve constructed by joining two half-ellipses by two parallel segments of equal length. The billiard inside it, as a map, generates a two parameters family of dynamical systems. It is known that the system is ergodic for a certain region of the parameter space. In this work we study the stability of a particular family of periodic orbits obtaining good bounds for the chaotic zone.
△ Less
Submitted 4 April, 1997;
originally announced April 1997.
-
Chaotic Properties of the Elliptical Stadium
Authors:
Roberto Markarian,
Sylvie Oliffson Kamphorst,
Sonia Pinto de Carvalho
Abstract:
The elliptical stadium is a curve constructed by joining two half-ellipses, with half axes a>1 and b=1, by two straight segments of equal length 2h. In this work we find bounds on h, for a close to 1, to assure the positiveness of a Lyapunov exponent. We conclude that, for these values of a and h, the elliptical stadium billiard mapping is ergodic and has the K-property.
The elliptical stadium is a curve constructed by joining two half-ellipses, with half axes a>1 and b=1, by two straight segments of equal length 2h. In this work we find bounds on h, for a close to 1, to assure the positiveness of a Lyapunov exponent. We conclude that, for these values of a and h, the elliptical stadium billiard mapping is ergodic and has the K-property.
△ Less
Submitted 4 January, 1995;
originally announced January 1995.
