Worst Case Performance

Worst-case performance analysis aims to identify and mitigate the most detrimental outcomes of algorithms and systems, crucial for safety-critical applications. Current research focuses on developing methods to efficiently compute worst-case performance, particularly for distributed optimization algorithms and deep neural networks, often leveraging techniques like semidefinite programming and adversarial training. This research is vital for improving the reliability and robustness of various systems, ranging from machine learning models to approximate nearest neighbor search algorithms, ensuring dependable performance even under challenging conditions. The ultimate goal is to create more resilient and trustworthy systems across diverse domains.

Papers

March 18, 2024

Exploiting Agent Symmetries for Performance Analysis of Distributed Optimization Methods
Sebastien Colla, Julien M. Hendrickx
Performance Analysis Semidefinite Programming Distributed Optimization Performance Estimation Algorithm Performance Worst Case Performance

December 11, 2023

Compute-in-Memory based Neural Network Accelerators for Safety-Critical Systems: Worst-Case Scenarios and Protections
Zheyu Yan, Xiaobo Sharon Hu, Yiyu Shi
Worst Case DNN Accelerator Safety Critical System Neural Network Accelerator Compute in Memory Deep Neural Network Performance Worst Case Performance

October 29, 2023

Worst-case Performance of Popular Approximate Nearest Neighbor Search Implementations: Guarantees and Limitations
Piotr Indyk, Haike Xu
Nearest Neighbor Fundamental Limitation Formal Guarantee Approximate Nearest Neighbor Search Graph Based Approximate Nearest Neighbor Worst Case Performance

October 15, 2023

Worst-Case Analysis is Maximum-A-Posteriori Estimation
Hongjun Wu, Di Wang
Worst Case Fuzz Testing Worst Case Performance

March 23, 2023

Benchmarking the Reliability of Post-training Quantization: a Particular Focus on Worst-case Performance
Zhihang Yuan, Jiawei Liu, Jiaxiang Wu, Dawei Yang, Qiang Wu, Guangyu Sun, Wenyu Liu, Xinggang Wang, Bingzhe Wu
Deep Neural Network Post Training Quantization Special Emphasis Worst Case Performance

April 10, 2022

Worst-case Performance of Greedy Policies in Bandits with Imperfect Context Observations
Hongju Park, Mohamad Kazem Shirani Faradonbeh
Contextual Bandit Sequential Decision Bandit Algorithm Greedy Policy Worst Case Performance Context Observation

February 15, 2022

Holistic Adversarial Robustness of Deep Learning Models
Pin-Yu Chen, Sijia Liu
Deep Learning Model Adversarial Robustness Worst Case Performance

February 2, 2022

Probabilistically Robust Learning: Balancing Average- and Worst-case Performance
Alexander Robey, Luiz F. O. Chamon, George J. Pappas, Hamed Hassani
Native Robustness Adversarial Training Robust Bayesian Worst Case Robustness Probabilistic Robustness Average Performance Worst Case Performance

Worst Case Performance

Papers

Exploiting Agent Symmetries for Performance Analysis of Distributed Optimization Methods

Compute-in-Memory based Neural Network Accelerators for Safety-Critical Systems: Worst-Case Scenarios and Protections

Worst-case Performance of Popular Approximate Nearest Neighbor Search Implementations: Guarantees and Limitations

Worst-Case Analysis is Maximum-A-Posteriori Estimation

Benchmarking the Reliability of Post-training Quantization: a Particular Focus on Worst-case Performance

Worst-case Performance of Greedy Policies in Bandits with Imperfect Context Observations

Holistic Adversarial Robustness of Deep Learning Models

Probabilistically Robust Learning: Balancing Average- and Worst-case Performance