Silviu Borac - Academia.edu (original) (raw)

Papers by Silviu Borac

Cornell University - arXiv, Oct 6, 2022

From human physiology to environmental evolution, important processes in nature often exhibit mea... more From human physiology to environmental evolution, important processes in nature often exhibit meaningful and strong periodic or quasi-periodic changes. Due to their inherent label scarcity, learning useful representations for periodic tasks with limited or no supervision is of great benefit. Yet, existing self-supervised learning (SSL) methods overlook the intrinsic periodicity in data, and fail to learn representations that capture periodic or frequency attributes. In this paper, we present SimPer, a simple contrastive SSL regime for learning periodic information in data. To exploit the periodic inductive bias, SimPer introduces customized augmentations, feature similarity measures, and a generalized contrastive loss for learning efficient and robust periodic representations. Extensive experiments on common real-world tasks in human behavior analysis, environmental sensing, and healthcare domains verify the superior performance of SimPer compared to state-of-the-art SSL methods, highlighting its intriguing properties including better data efficiency, robustness to spurious correlations, and generalization to distribution shifts.

The integral equation arising from space variant 2-D texture resampling is reformulated through w... more The integral equation arising from space variant 2-D texture resampling is reformulated through wavelet analysis. We transform the standard convolution integral in texture space into an inner product over sparse representations for both the texture and the warped filter function. This yields an algorithm that operates in constant time in the area of the domain of convolution, and that is sensitive to the frequency content of both the filter and the texture. The reformulation admits further acceleration for space-invariant resampling. 1 Issues The problem of efficiently processing two-dimensional textures that undergo space variant warps is both important and technically challenging. In most rendering situations, textures are differentially mapped onto surfaces through parameterisation and they are viewed through a perspective projection. Both mappings in principle require texture resampling operations at a cost that can vary significantly from pixel to pixel. The constant-time space...

For many years the structure of 1+1 dimensional systems (one space + one time dimension) has been... more For many years the structure of 1+1 dimensional systems (one space + one time dimension) has been of great interest for mathematical physics. The Bethe Ansatz is a unifying method connecting a wide range of different problems in these areas. It has been applied to magnetic chains in solid state physics as well as to problems of (relativistic) field theory and elementary particle physics (the Sine-Gordon equation, solitons, integrable models). After 20 years of experience in this field we have the feeling that 1+1 D systems are of importance for 4-dimensional quantum field theory too. In his last address to the International Conference Feynman mentions the Bethe Ansatz and advocates the picture that high energy scattering is mainly 1+1 dimensional with finitely many transversal modes only. An application to gravity and gauge theory has been found by Verlinde et al.[1], when they separated the 4D Einstein equations into two sets of 2D equations. In general we believe that the search f...

medRxiv, 2021

Measuring vital signs plays a key role in both patient care and wellness, but can be challenging ... more Measuring vital signs plays a key role in both patient care and wellness, but can be challenging outside of medical settings due to the lack of specialized equipment. In this study, we prospectively evaluated smartphone camera-based techniques for measuring heart rate (HR) and respiratory rate (RR) for consumer wellness use. HR was measured by placing the finger over the rear-facing camera, while RR was measured via a video of the participants sitting still in front of the front-facing camera. In the HR study of 95 participants (with a protocol that included both measurements at rest and post exercise), the mean absolute percent error (MAPE) ± standard deviation of the measurement was 1.6% ± 4.3%, which was significantly lower than the pre-specified goal of 5%. No significant differences in the MAPE were present across colorimeter-measured skin-tone subgroups: 1.8% ± 4.5% for very light to intermediate, 1.3% ± 3.3% for tan and brown, and 1.8% ± 4.9% for dark. In the RR study of 50 p...

We present a performance optimization study for an incompressible fluid simulation code designed ... more We present a performance optimization study for an incompressible fluid simulation code designed for animated special effects. The solver uses a marker-and-cell (MAC) method to simulate the motion of liquids and gases. The main challenges in the optimization were identifying efficient data structures for managing the particle data, and rewriting the iterative pressure projection step so as to introduce threading without compromising the convergence rate. On average we achieved an improvement of over 300% in the performance of the serial implementation and parallel efficiency of 65% on an 8-core desktop by threading portions of the application.

Proceedings of the ACM on Computer Graphics and Interactive Techniques

We present a new algorithm for image-based simplification of complex scenes for virtual reality (... more We present a new algorithm for image-based simplification of complex scenes for virtual reality (VR). The algorithm transforms geometrically-detailed environments into a layered quad tile representation that is optimized for a specified viewing region and is renderable on low-power mobile-class VR devices. A novel constrained optimization formulation ensures that the scene can be rendered within a predetermined compute budget, with limits on both primitive count and fill rate. Furthermore, we introduce a new method for texturing from point samples of the original scene geometry that generates high-quality silhouettes without the drawbacks of traditional point splatting. The resulting representation achieves a visual fidelity that was previously impossible on mobile graphics hardware; our algorithm can typically generate a high-quality representation of visually-rich scenes with billions of triangles using just 72k triangles and a single high-resolution texture map (with generally on...

Proceedings of the 2015 Symposium on Digital Production, 2015

We present a hybrid direct and projective iterative algorithm for simulating hair. The algorithm ... more We present a hybrid direct and projective iterative algorithm for simulating hair. The algorithm is fast, stable, visually accurate, uses a comprehensive elasticity model with a well defined continuum limit. The low resolution simulation is a good predictor of higher resolution results which we attribute to our novel treatment of bending and twisting forces. Timesteps may be as large a single step for each frame of 24 Hz animation. The treatment of collisions is implicit and does not introduce any artificial strain or forces.

Freeform Tools in CAD Systems, 1991

Although currently available CAD systems excel at producing geometry, many users require higher q... more Although currently available CAD systems excel at producing geometry, many users require higher quality visualization of their data than obtainable from the CAD system itself. The visualization problem may be decomposed into (1) accepting input data, upon which operations such as the detection of adjacent surfaces may need to be performed, (2) conversion of free-form surfaces into a form acceptable to the Tenderer, (3) defining the scene, and (4) rendering the image. Issues and solutions to each of these problems are discussed in the context of research and development at mental images, including the adjacency detection and edge merging problems, triangulation of free-form surfaces in parameter space under various constraints in three dimensional space and use of implicit (algebraic) surfaces as an alternative free-form surface visualization technique.

Eurographics, 1996

The integral equation arising from space variant 2-D texture resampling is reformulated through w... more The integral equation arising from space variant 2-D texture resampling is reformulated through wavelet analysis. We transform the standard convolution integral in texture space into an inner product over sparse representations for both the texture and the warped filter function. This yields an algorithm that operates in constant time in the area of the domain of convolution, and that is sensitive to the frequency content of both the filter and the texture. The reformulation admits further acceleration for space-invariant resampling.

Journal of Mathematical Physics, 1995

The von Neumann algebra generated by a pair of projections on a separable Hilbert space occurs na... more The von Neumann algebra generated by a pair of projections on a separable Hilbert space occurs naturally in the study of quantum Hall systems. By a systematic use of Kato’s dual pair of self‐adjoint operators corresponding to the two projections, the argument referring to the structure of the algebra can be simplified and the existence of an action of the group SU(2) on the algebra via inner automorphisms can be shown. Also, the algebra can be given an r‐structure with respect to which it is r‐commutative, and an extension of the classical (commutative) differential calculus becomes possible.

Cornell University - arXiv, Oct 6, 2022

From human physiology to environmental evolution, important processes in nature often exhibit mea... more From human physiology to environmental evolution, important processes in nature often exhibit meaningful and strong periodic or quasi-periodic changes. Due to their inherent label scarcity, learning useful representations for periodic tasks with limited or no supervision is of great benefit. Yet, existing self-supervised learning (SSL) methods overlook the intrinsic periodicity in data, and fail to learn representations that capture periodic or frequency attributes. In this paper, we present SimPer, a simple contrastive SSL regime for learning periodic information in data. To exploit the periodic inductive bias, SimPer introduces customized augmentations, feature similarity measures, and a generalized contrastive loss for learning efficient and robust periodic representations. Extensive experiments on common real-world tasks in human behavior analysis, environmental sensing, and healthcare domains verify the superior performance of SimPer compared to state-of-the-art SSL methods, highlighting its intriguing properties including better data efficiency, robustness to spurious correlations, and generalization to distribution shifts.

The integral equation arising from space variant 2-D texture resampling is reformulated through w... more The integral equation arising from space variant 2-D texture resampling is reformulated through wavelet analysis. We transform the standard convolution integral in texture space into an inner product over sparse representations for both the texture and the warped filter function. This yields an algorithm that operates in constant time in the area of the domain of convolution, and that is sensitive to the frequency content of both the filter and the texture. The reformulation admits further acceleration for space-invariant resampling. 1 Issues The problem of efficiently processing two-dimensional textures that undergo space variant warps is both important and technically challenging. In most rendering situations, textures are differentially mapped onto surfaces through parameterisation and they are viewed through a perspective projection. Both mappings in principle require texture resampling operations at a cost that can vary significantly from pixel to pixel. The constant-time space...

For many years the structure of 1+1 dimensional systems (one space + one time dimension) has been... more For many years the structure of 1+1 dimensional systems (one space + one time dimension) has been of great interest for mathematical physics. The Bethe Ansatz is a unifying method connecting a wide range of different problems in these areas. It has been applied to magnetic chains in solid state physics as well as to problems of (relativistic) field theory and elementary particle physics (the Sine-Gordon equation, solitons, integrable models). After 20 years of experience in this field we have the feeling that 1+1 D systems are of importance for 4-dimensional quantum field theory too. In his last address to the International Conference Feynman mentions the Bethe Ansatz and advocates the picture that high energy scattering is mainly 1+1 dimensional with finitely many transversal modes only. An application to gravity and gauge theory has been found by Verlinde et al.[1], when they separated the 4D Einstein equations into two sets of 2D equations. In general we believe that the search f...

medRxiv, 2021

Measuring vital signs plays a key role in both patient care and wellness, but can be challenging ... more Measuring vital signs plays a key role in both patient care and wellness, but can be challenging outside of medical settings due to the lack of specialized equipment. In this study, we prospectively evaluated smartphone camera-based techniques for measuring heart rate (HR) and respiratory rate (RR) for consumer wellness use. HR was measured by placing the finger over the rear-facing camera, while RR was measured via a video of the participants sitting still in front of the front-facing camera. In the HR study of 95 participants (with a protocol that included both measurements at rest and post exercise), the mean absolute percent error (MAPE) ± standard deviation of the measurement was 1.6% ± 4.3%, which was significantly lower than the pre-specified goal of 5%. No significant differences in the MAPE were present across colorimeter-measured skin-tone subgroups: 1.8% ± 4.5% for very light to intermediate, 1.3% ± 3.3% for tan and brown, and 1.8% ± 4.9% for dark. In the RR study of 50 p...

We present a performance optimization study for an incompressible fluid simulation code designed ... more We present a performance optimization study for an incompressible fluid simulation code designed for animated special effects. The solver uses a marker-and-cell (MAC) method to simulate the motion of liquids and gases. The main challenges in the optimization were identifying efficient data structures for managing the particle data, and rewriting the iterative pressure projection step so as to introduce threading without compromising the convergence rate. On average we achieved an improvement of over 300% in the performance of the serial implementation and parallel efficiency of 65% on an 8-core desktop by threading portions of the application.

Proceedings of the ACM on Computer Graphics and Interactive Techniques

We present a new algorithm for image-based simplification of complex scenes for virtual reality (... more We present a new algorithm for image-based simplification of complex scenes for virtual reality (VR). The algorithm transforms geometrically-detailed environments into a layered quad tile representation that is optimized for a specified viewing region and is renderable on low-power mobile-class VR devices. A novel constrained optimization formulation ensures that the scene can be rendered within a predetermined compute budget, with limits on both primitive count and fill rate. Furthermore, we introduce a new method for texturing from point samples of the original scene geometry that generates high-quality silhouettes without the drawbacks of traditional point splatting. The resulting representation achieves a visual fidelity that was previously impossible on mobile graphics hardware; our algorithm can typically generate a high-quality representation of visually-rich scenes with billions of triangles using just 72k triangles and a single high-resolution texture map (with generally on...

Proceedings of the 2015 Symposium on Digital Production, 2015

We present a hybrid direct and projective iterative algorithm for simulating hair. The algorithm ... more We present a hybrid direct and projective iterative algorithm for simulating hair. The algorithm is fast, stable, visually accurate, uses a comprehensive elasticity model with a well defined continuum limit. The low resolution simulation is a good predictor of higher resolution results which we attribute to our novel treatment of bending and twisting forces. Timesteps may be as large a single step for each frame of 24 Hz animation. The treatment of collisions is implicit and does not introduce any artificial strain or forces.

Freeform Tools in CAD Systems, 1991

Although currently available CAD systems excel at producing geometry, many users require higher q... more Although currently available CAD systems excel at producing geometry, many users require higher quality visualization of their data than obtainable from the CAD system itself. The visualization problem may be decomposed into (1) accepting input data, upon which operations such as the detection of adjacent surfaces may need to be performed, (2) conversion of free-form surfaces into a form acceptable to the Tenderer, (3) defining the scene, and (4) rendering the image. Issues and solutions to each of these problems are discussed in the context of research and development at mental images, including the adjacency detection and edge merging problems, triangulation of free-form surfaces in parameter space under various constraints in three dimensional space and use of implicit (algebraic) surfaces as an alternative free-form surface visualization technique.

Eurographics, 1996

The integral equation arising from space variant 2-D texture resampling is reformulated through w... more The integral equation arising from space variant 2-D texture resampling is reformulated through wavelet analysis. We transform the standard convolution integral in texture space into an inner product over sparse representations for both the texture and the warped filter function. This yields an algorithm that operates in constant time in the area of the domain of convolution, and that is sensitive to the frequency content of both the filter and the texture. The reformulation admits further acceleration for space-invariant resampling.

Journal of Mathematical Physics, 1995

The von Neumann algebra generated by a pair of projections on a separable Hilbert space occurs na... more The von Neumann algebra generated by a pair of projections on a separable Hilbert space occurs naturally in the study of quantum Hall systems. By a systematic use of Kato’s dual pair of self‐adjoint operators corresponding to the two projections, the argument referring to the structure of the algebra can be simplified and the existence of an action of the group SU(2) on the algebra via inner automorphisms can be shown. Also, the algebra can be given an r‐structure with respect to which it is r‐commutative, and an extension of the classical (commutative) differential calculus becomes possible.