Deep Learning Architectures for Robust Classification Under Adversarial Noise

Abstract

This report focuses on the problem of designing robust classifiers to images that are distorted by noise. The approach taken was robust optimization where the goal was to optimize in the worst case over a class of objective functions. A theoretical framework with strong guarantees was developed. In particular it was shown that given a classifier that has alpha accuracy over a finite number of attacks, one can develop a robust classifier that is an arbitrarily close to be an alpha approximation to the optimal robust classifier. These results were applied to robust neural network training and approach was evaluated experimentally on corrupted character classification.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Aug 01, 2019
Accession Number
AD1078373

Entities

People

  • Yaron Singer

Organizations

  • Harvard University

Tags

Communities of Interest

  • Sensors
  • Space

DTIC Thesaurus Topics

  • Accuracy
  • Air Force
  • Air Force Research Laboratories
  • Algorithms
  • Artificial Intelligence Software
  • Classification
  • Data Sets
  • Deep Learning
  • Government Procurement
  • Governments
  • Guarantees
  • Learning
  • Machine Learning
  • Neural Networks
  • Optimization
  • Standards
  • Training

Fields of Study

  • Computer science

Readers

  • Computer Vision.
  • Distributed Systems and Data Platform Development
  • Statistical inference.

Technology Areas

  • AI & ML
  • AI & ML - Machine Learning Algorithms
  • AI & ML - Neural Networks