Native Robustness

Native robustness in machine learning focuses on developing models inherently resistant to various forms of input perturbations, including adversarial attacks and noisy data, without relying solely on post-hoc defenses. Current research emphasizes techniques like ensemble methods, reprogramming existing models, and modifying training procedures (e.g., using different learning rates for specific model layers or incorporating regularization methods) to improve robustness across diverse model architectures, including convolutional neural networks, vision transformers, and large language models. This field is crucial for deploying reliable AI systems in safety-critical applications, such as healthcare and autonomous driving, where model resilience to unexpected inputs is paramount.

Papers

August 8, 2023

Comprehensive Assessment of the Performance of Deep Learning Classifiers Reveals a Surprising Lack of Robustness
Michael W. Spratling
Native Robustness System Performance Comprehensive Evaluation Robustness Evaluation Robust Evaluation Deep Learning Classifier Benchmark Score Robust Machine Learning Lack Thereof State of the Art Robustness

August 7, 2023

Distributionally Robust Classification on a Data Budget
Benjamin Feuer, Ameya Joshi, Minh Pham, Chinmay Hegde
Native Robustness Robust Learning Distributionally Robust Distributional Robustness Robustness Gap Data Budgeting

August 4, 2023

August 1, 2023

July 31, 2023

July 29, 2023

Evaluating the Robustness of Test Selection Methods for Deep Neural Networks
Qiang Hu, Yuejun Guo, Xiaofei Xie, Maxime Cordy, Wei Ma, Mike Papadakis, Yves Le Traon
Deep Neural Network Native Robustness Fault Detection Test Data Test Selection

July 28, 2023

Worrisome Properties of Neural Network Controllers and Their Symbolic Representations
Jacek Cyranka, Kevin E M Church, Jean-Philippe Lessard
Native Robustness Neural Network Controller Reinforcement Learning Benchmark Symbolic Representation Robustness Analysis

July 27, 2023

A/B Testing and Best-arm Identification for Linear Bandits with Robustness to Non-stationarity
Zhihan Xiong, Romain Camilleri, Maryam Fazel, Lalit Jain, Kevin Jamieson
Native Robustness Linear Bandit Best Arm Identification Optimal Design Non Stationarity Fixed Budget Best Arm Identification B Test

July 26, 2023

July 25, 2023

July 24, 2023

Native Robustness

Papers

Comprehensive Assessment of the Performance of Deep Learning Classifiers Reveals a Surprising Lack of Robustness

Distributionally Robust Classification on a Data Budget

SureFED: Robust Federated Learning via Uncertainty-Aware Inward and Outward Inspection

RobustMQ: Benchmarking Robustness of Quantized Models

Training on Foveated Images Improves Robustness to Adversarial Attacks

Robust Linear Regression: Phase-Transitions and Precise Tradeoffs for General Norms

Is Last Layer Re-Training Truly Sufficient for Robustness to Spurious Correlations?

Improving Generalization of Adversarial Training via Robust Critical Fine-Tuning

Towards Trustworthy and Aligned Machine Learning: A Data-centric Survey with Causality Perspectives

Probabilistically robust conformal prediction

Evaluating the Robustness of Test Selection Methods for Deep Neural Networks

Worrisome Properties of Neural Network Controllers and Their Symbolic Representations

A/B Testing and Best-arm Identification for Linear Bandits with Robustness to Non-stationarity

Robustness Verification of Deep Neural Networks using Star-Based Reachability Analysis with Variable-Length Time Series Input

Characterizing Data Point Vulnerability via Average-Case Robustness

Scaff-PD: Communication Efficient Fair and Robust Federated Learning

Federated Distributionally Robust Optimization with Non-Convex Objectives: Algorithm and Analysis

Investigating the Robustness of Sequential Recommender Systems Against Training Data Perturbations

Adaptive Certified Training: Towards Better Accuracy-Robustness Tradeoffs

On the Connection between Pre-training Data Diversity and Fine-tuning Robustness