Dhruv V. Patel
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Dhruv V. Patel has a total of 23 co-authors. Here are the top co-authors by number of shared publications:
- Assad A. Oberai
Department of Aerospace and Mechanical Engineering, University of Southern California, Los Angeles, CA 90089, USA
6 shared publications Assad A. Oberai
4 shared publications- Deep Ray
University of Maryland
4 shared publications Deep Ray
2 shared publications- Aakash Aanegola1 shared publication
Adriana Vega
1 shared publication
Latest Publications
- 1 January 2024
- Published by American Institute of Mathematical Sciences (AIMS) in Numerical Algebra, Control & Optimization
- Vol. 14 (1) , 160-189
The efficacy and generalizability of conditional GANs for posterior inference in physics-based inverse problems
In this work, we train conditional Wasserstein generative adversarial networks to effectively sample from the posterior of physics-based Bayesian inference problems. The generator is constructed using a U-Net architecture, with the latent information injected using conditional instance normalization. The former facilitates a multiscale inverse map, while the latter enables the decoupling of the latent space dimension from the dimension of the measurement, and introduces stochasticity at all scales of the U-Net. We solve PDE-based inverse problems to demonstrate the performance of our approach in quantifying the uncertainty in the inferred field. Further, we show the generator can learn inverse maps which are local in nature, which in turn promotes generalizability when testing with out-of-distribution samples.
- 1 January 2024
- Published by American Institute of Mathematical Sciences (AIMS) in Numerical Algebra, Control & Optimization
- Vol. 14 (1) , 160-189
The efficacy and generalizability of conditional GANs for posterior inference in physics-based inverse problems
In this work, we train conditional Wasserstein generative adversarial networks to effectively sample from the posterior of physics-based Bayesian inference problems. The generator is constructed using a U-Net architecture, with the latent information injected using conditional instance normalization. The former facilitates a multiscale inverse map, while the latter enables the decoupling of the latent space dimension from the dimension of the measurement, and introduces stochasticity at all scales of the U-Net. We solve PDE-based inverse problems to demonstrate the performance of our approach in quantifying the uncertainty in the inferred field. Further, we show the generator can learn inverse maps which are local in nature, which in turn promotes generalizability when testing with out-of-distribution samples.
- 18 November 2023
- Published by Elsevier in Computer Methods in Applied Mechanics and Engineering
- Vol. 419
Variationally mimetic operator networks
- 18 November 2023
- Published by Elsevier in Computer Methods in Applied Mechanics and Engineering
- Vol. 419
Variationally mimetic operator networks
- Open Access
Variationally Mimetic Operator Networks
In recent years operator networks have emerged as promising deep learning tools for approximating the solution to partial differential equations (PDEs). These networks map input functions that describe material properties, forcing functions and boundary data to the solution of a PDE. This work describes a new architecture for operator networks that mimics the form of the numerical solution obtained from an approximate variational or weak formulation of the problem. The application of these ideas to a generic elliptic PDE leads to a variationally mimetic operator network (VarMiON). Like the conventional Deep Operator Network (DeepONet) the VarMiON is also composed of a sub-network that constructs the basis functions for the output and another that constructs the coefficients for these basis functions. However, in contrast to the DeepONet, the architecture of these sub-networks in the VarMiON is precisely determined. An analysis of the error in the VarMiON solution reveals that it contains contributions from the error in the training data, the training error, the quadrature error in sampling input and output functions, and a "covering error" that measures the distance between the test input functions and the nearest functions in the training dataset. It also depends on the stability constants for the exact solution operator and its VarMiON approximation. The application of the VarMiON to a canonical elliptic PDE and a nonlinear PDE reveals that for approximately the same number of network parameters, on average the VarMiON incurs smaller errors than a standard DeepONet and a recently proposed multiple-input operator network (MIONet). Further, its performance is more robust to variations in input functions, the techniques used to sample the input and output functions, the techniques used to construct the basis functions, and the number of input functions.
Open AccessVariationally Mimetic Operator Networks
In recent years operator networks have emerged as promising deep learning tools for approximating the solution to partial differential equations (PDEs). These networks map input functions that describe material properties, forcing functions and boundary data to the solution of a PDE. This work describes a new architecture for operator networks that mimics the form of the numerical solution obtained from an approximate variational or weak formulation of the problem. The application of these ideas to a generic elliptic PDE leads to a variationally mimetic operator network (VarMiON). Like the conventional Deep Operator Network (DeepONet) the VarMiON is also composed of a sub-network that constructs the basis functions for the output and another that constructs the coefficients for these basis functions. However, in contrast to the DeepONet, the architecture of these sub-networks in the VarMiON is precisely determined. An analysis of the error in the VarMiON solution reveals that it contains contributions from the error in the training data, the training error, the quadrature error in sampling input and output functions, and a "covering error" that measures the distance between the test input functions and the nearest functions in the training dataset. It also depends on the stability constants for the exact solution operator and its VarMiON approximation. The application of the VarMiON to a canonical elliptic PDE and a nonlinear PDE reveals that for approximately the same number of network parameters, on average the VarMiON incurs smaller errors than a standard DeepONet and a recently proposed multiple-input operator network (MIONet). Further, its performance is more robust to variations in input functions, the techniques used to sample the input and output functions, the techniques used to construct the basis functions, and the number of input functions.
- Open Access
The efficacy and generalizability of conditional GANs for posterior inference in physics-based inverse problems
In this work, we train conditional Wasserstein generative adversarial networks to effectively sample from the posterior of physics-based Bayesian inference problems. The generator is constructed using a U-Net architecture, with the latent information injected using conditional instance normalization. The former facilitates a multiscale inverse map, while the latter enables the decoupling of the latent space dimension from the dimension of the measurement, and introduces stochasticity at all scales of the U-Net. We solve PDE-based inverse problems to demonstrate the performance of our approach in quantifying the uncertainty in the inferred field. Further, we show the generator can learn inverse maps which are local in nature, which in turn promotes generalizability when testing with out-of-distribution samples.
Open AccessThe efficacy and generalizability of conditional GANs for posterior inference in physics-based inverse problems
In this work, we train conditional Wasserstein generative adversarial networks to effectively sample from the posterior of physics-based Bayesian inference problems. The generator is constructed using a U-Net architecture, with the latent information injected using conditional instance normalization. The former facilitates a multiscale inverse map, while the latter enables the decoupling of the latent space dimension from the dimension of the measurement, and introduces stochasticity at all scales of the U-Net. We solve PDE-based inverse problems to demonstrate the performance of our approach in quantifying the uncertainty in the inferred field. Further, we show the generator can learn inverse maps which are local in nature, which in turn promotes generalizability when testing with out-of-distribution samples.
- Open Access
GDIP: Gated Differentiable Image Processing for Object-Detection in Adverse Conditions
Detecting objects under adverse weather and lighting conditions is crucial for the safe and continuous operation of an autonomous vehicle, and remains an unsolved problem. We present a Gated Differentiable Image Processing (GDIP) block, a domain-agnostic network architecture, which can be plugged into existing object detection networks (e.g., Yolo) and trained end-to-end with adverse condition images such as those captured under fog and low lighting. Our proposed GDIP block learns to enhance images directly through the downstream object detection loss. This is achieved by learning parameters of multiple image pre-processing (IP) techniques that operate concurrently, with their outputs combined using weights learned through a novel gating mechanism. We further improve GDIP through a multi-stage guidance procedure for progressive image enhancement. Finally, trading off accuracy for speed, we propose a variant of GDIP that can be used as a regularizer for training Yolo, which eliminates the need for GDIP-based image enhancement during inference, resulting in higher throughput and plausible real-world deployment. We demonstrate significant improvement in detection performance over several state-of-the-art methods through quantitative and qualitative studies on synthetic datasets such as PascalVOC, and real-world foggy (RTTS) and low-lighting (ExDark) datasets.
Open AccessGDIP: Gated Differentiable Image Processing for Object-Detection in Adverse Conditions
Detecting objects under adverse weather and lighting conditions is crucial for the safe and continuous operation of an autonomous vehicle, and remains an unsolved problem. We present a Gated Differentiable Image Processing (GDIP) block, a domain-agnostic network architecture, which can be plugged into existing object detection networks (e.g., Yolo) and trained end-to-end with adverse condition images such as those captured under fog and low lighting. Our proposed GDIP block learns to enhance images directly through the downstream object detection loss. This is achieved by learning parameters of multiple image pre-processing (IP) techniques that operate concurrently, with their outputs combined using weights learned through a novel gating mechanism. We further improve GDIP through a multi-stage guidance procedure for progressive image enhancement. Finally, trading off accuracy for speed, we propose a variant of GDIP that can be used as a regularizer for training Yolo, which eliminates the need for GDIP-based image enhancement during inference, resulting in higher throughput and plausible real-world deployment. We demonstrate significant improvement in detection performance over several state-of-the-art methods through quantitative and qualitative studies on synthetic datasets such as PascalVOC, and real-world foggy (RTTS) and low-lighting (ExDark) datasets.
- 5 September 2022
- Published by Elsevier in Computer Methods in Applied Mechanics and Engineering
- Vol. 400
Solution of physics-based Bayesian inverse problems with deep generative priors
- 5 September 2022
- Published by Elsevier in Computer Methods in Applied Mechanics and Engineering
- Vol. 400
Solution of physics-based Bayesian inverse problems with deep generative priors
- 1 January 2021
- Published by Society for Industrial & Applied Mathematics (SIAM) in SIAM/ASA Journal on Uncertainty Quantification
- Vol. 9 (3) , 1314-1343
GAN-Based Priors for Quantifying Uncertainty in Supervised Learning
Machine learning and deep learning algorithms are frequently used in critical tasks where their output is used in high-stakes downstream applications. In these cases it is important to quantify the...
- 1 January 2021
- Published by Society for Industrial & Applied Mathematics (SIAM) in SIAM/ASA Journal on Uncertainty Quantification
- Vol. 9 (3) , 1314-1343
GAN-Based Priors for Quantifying Uncertainty in Supervised Learning
Machine learning and deep learning algorithms are frequently used in critical tasks where their output is used in high-stakes downstream applications. In these cases it is important to quantify the...
- 1 April 2020
- Published by SPIE-Intl Soc Optical Eng in Journal of Micro/Nanolithography, MEMS, and MOEMS
- Vol. 19 (02) , 024801-15
Deep learning-based detection, classification, and localization of defects in semiconductor processes
Defects in semiconductor processes can limit yield, increase overall production cost, and also lead to time-dependent critical component failures. Current state-of-the-art optical and electron beam (EB) inspection systems rely on rule-based techniques for defect detection and classification, which are usually rigid in their comparative processes. This rigidity limits overall capability and increases relative engineering time to classify nuisance defects. This is further challenged due to shrinkage of pattern dimensions for advanced nodes. We propose a deep learning-based workflow that circumvents these challenges and enables accurate defect detection, classification, and localization in a unified framework. In particular, we train convolutional neural network-based models using high-resolution EB images of wafers patterned with various types of intentional defects and achieve robust defect detection and classification performance. Furthermore, we generate class activation maps to demonstrate defect localization capability of the model “without” explicitly training it with defect location information. To understand the underlying decision-making process of these deep models, we analyze the learned filters in pixel space and Fourier space and interpret the various operations at different layers. We achieve high sensitivity (97%) and specificity (100%) along with rapid and accurate defect localization. We also test performance of the proposed workflow on images from two distinct patterns and find that in order to retain high accuracy a modest level of retraining is necessary.
- 1 April 2020
- Published by SPIE-Intl Soc Optical Eng in Journal of Micro/Nanolithography, MEMS, and MOEMS
- Vol. 19 (02) , 024801-15
Deep learning-based detection, classification, and localization of defects in semiconductor processes
Defects in semiconductor processes can limit yield, increase overall production cost, and also lead to time-dependent critical component failures. Current state-of-the-art optical and electron beam (EB) inspection systems rely on rule-based techniques for defect detection and classification, which are usually rigid in their comparative processes. This rigidity limits overall capability and increases relative engineering time to classify nuisance defects. This is further challenged due to shrinkage of pattern dimensions for advanced nodes. We propose a deep learning-based workflow that circumvents these challenges and enables accurate defect detection, classification, and localization in a unified framework. In particular, we train convolutional neural network-based models using high-resolution EB images of wafers patterned with various types of intentional defects and achieve robust defect detection and classification performance. Furthermore, we generate class activation maps to demonstrate defect localization capability of the model “without” explicitly training it with defect location information. To understand the underlying decision-making process of these deep models, we analyze the learned filters in pixel space and Fourier space and interpret the various operations at different layers. We achieve high sensitivity (97%) and specificity (100%) along with rapid and accurate defect localization. We also test performance of the proposed workflow on images from two distinct patterns and find that in order to retain high accuracy a modest level of retraining is necessary.
- 1 August 2019
- Published by Elsevier in Computer Methods in Applied Mechanics and Engineering
- Vol. 353, 448-466
Circumventing the solution of inverse problems in mechanics through deep learning: Application to elasticity imaging
- 1 August 2019
- Published by Elsevier in Computer Methods in Applied Mechanics and Engineering
- Vol. 353, 448-466
Circumventing the solution of inverse problems in mechanics through deep learning: Application to elasticity imaging