Differentiable Surrogate

Differentiable surrogates are approximations of complex, often non-differentiable, functions or models (e.g., black-box simulators, optimization problems) that allow for efficient gradient-based optimization. Current research focuses on developing and applying differentiable surrogates using neural networks, particularly for tasks like hyperparameter optimization, distributionally robust optimization, and improving the efficiency of existing algorithms (e.g., Model Predictive Control). This approach significantly enhances the tractability of otherwise computationally expensive or intractable problems across diverse fields, from image processing and building optimization to particle physics simulations. The resulting improvements in optimization speed and accuracy have broad implications for scientific modeling and engineering applications.

Papers

July 23, 2024

Learning Instance-Specific Parameters of Black-Box Models Using Differentiable Surrogates
Arnisha Khondaker, Nilanjan Ray
LeArning Abstract Black Box Hyperparameter Optimization Black Box Model Instance Level Differentiable Surrogate

June 24, 2024

Differentiable Distributionally Robust Optimization Layers
Xutao Ma, Chao Ning, Wenli Du
Robust Optimization Distributionally Robust Optimization Ambiguity Set Wasserstein Uncertainty Differentiable Surrogate

June 6, 2024

Simulating, Fast and Slow: Learning Policies for Black-Box Optimization
Fabio Valerio Massoli, Tim Bakker, Thomas Hehn, Tribhuvanesh Orekondy, Arash Behboodi
Optimization Purpose Simulation Study Black Box Optimization Numerical Optimization Black Box Optimization Problem Learning Policy Differentiable Surrogate Local Surrogate

January 13, 2024

Exploring Adversarial Attacks against Latent Diffusion Model from the Perspective of Adversarial Transferability
Junxi Chen, Junhao Dong, Xiaohua Xie
Adversarial Attack Adversarial Example Latent Diffusion Model Visual Perspective Differentiable Surrogate

October 23, 2023

Theoretically Grounded Loss Functions and Algorithms for Score-Based Multi-Class Abstention
Anqi Mao, Mehryar Mohri, Yutao Zhong
Practical Algorithm Loss Function Surrogate Loss Differentiable Surrogate

January 31, 2023

Benchmarking Model Predictive Control Algorithms in Building Optimization Testing Framework (BOPTEST)
Saman Mostafavi, Chihyeon Song, Aayushman Sharma, Raman Goyal, Alejandro Brito
Predictive Control Model Predictive Control Physical Simulation Building Energy Predictive Controller Differentiable Surrogate

November 8, 2022

Theoretical analysis and experimental validation of volume bias of soft Dice optimized segmentation maps in the context of inherent uncertainty
Jeroen Bertels, David Robben, Dirk Vandermeulen, Paul Suetens
Experimental Validation Automatic Volume Segmentation Map Differentiable Surrogate Dice Similarity Coefficient Inherent Uncertainty

November 2, 2022

Implicit Neural Representation as a Differentiable Surrogate for Photon Propagation in a Monolithic Neutrino Detector
Minjie Lei, Ka Vang Tsang, Sean Gasiorowski, Chuan Li, Youssef Nashed, Gianluca Petrillo, Olivia Piazza, Daniel Ratner, Kazuhiro Terao
Implicit Neural Representation Radiation Detector Detector Simulation Differentiable Surrogate Light Propagation Sinusoidal Representation Network Neutrino Detector

June 7, 2022

Combining Genetic Programming and Particle Swarm Optimization to Simplify Rugged Landscapes Exploration
Gloria Pietropolli, Giuliamaria Menara, Mauro Castelli
Particle Swarm Optimization Genetic Programming Metaheuristic Algorithm Differentiable Surrogate

May 20, 2022

A Case of Exponential Convergence Rates for SVM
Vivien Cabannes, Stefano Vigogna
Case Relevance Convergence Rate Generalization Guarantee Exponential Convergence Rate Support Vector Differentiable Surrogate

Differentiable Surrogate

Papers

Learning Instance-Specific Parameters of Black-Box Models Using Differentiable Surrogates

Differentiable Distributionally Robust Optimization Layers

Simulating, Fast and Slow: Learning Policies for Black-Box Optimization

Exploring Adversarial Attacks against Latent Diffusion Model from the Perspective of Adversarial Transferability

Theoretically Grounded Loss Functions and Algorithms for Score-Based Multi-Class Abstention

Benchmarking Model Predictive Control Algorithms in Building Optimization Testing Framework (BOPTEST)

Theoretical analysis and experimental validation of volume bias of soft Dice optimized segmentation maps in the context of inherent uncertainty

Implicit Neural Representation as a Differentiable Surrogate for Photon Propagation in a Monolithic Neutrino Detector

Combining Genetic Programming and Particle Swarm Optimization to Simplify Rugged Landscapes Exploration

A Case of Exponential Convergence Rates for SVM