Published Feb 14, 2020

Adversarial Explanations

Dive into the complexities of neural network interpretability and robustness with Walt Woods as he unpacks adversarial explanations, explores the Leapshift constraint, and reveals the vulnerabilities of adversarial examples in image classification systems, challenging the black box nature of AI models.

Episode Highlights

Topics covered

Episode Highlights

Related Episodes

Defending Against Adversarial Attacks
Answers 383 questions
Robustness to Unforeseen Adversarial Attacks
Answers 383 questions
Explainable Climate Science
Answers 383 questions
Self-Explaining AI
Answers 383 questions
Annotator Bias
Answers 383 questions
Machine Learning Done Wrong
Answers 383 questions
Explainable K-Means
Answers 383 questions
Visualization and Interpretability
Answers 383 questions
Fooling Computer Vision
Answers 383 questions
Understanding Neural Networks
Answers 383 questions
Interpretability Practitioners
Answers 383 questions
Authorship Attribution
Answers 383 questions
Fake News Midterm
Answers 383 questions
Easily Fooling Deep Neural Networks
Answers 383 questions
Multi-Agent Diverse Generative Adversarial Networks
Answers 383 questions

Dexa/Kyle Polich

Adversarial Explanations

Topics covered

Popular Clips

Exploring Adversarial Explanations

Alarming Biases

Unlocking Neural Network Parameters

Adversarial Examples

Robust Classifiers, Domains, and Training Loss

Optimizing Neural Networks

Uncovering Data Myths

Implementing Half Huber Rectified Linear Unit

Adversarial Explanations

Trust in Critical Systems

Evaluating Classifier Behavior

Improved Explanations

Adversarial Explanations

Accuracy Robustness Area

Episode Highlights

Neural Network Explanations

Improving Network Robustness

Adversarial Examples

Related Episodes

Defending Against Adversarial Attacks

Robustness to Unforeseen Adversarial Attacks

Explainable Climate Science

Self-Explaining AI

Annotator Bias

Machine Learning Done Wrong

Explainable K-Means

Visualization and Interpretability

Fooling Computer Vision

Understanding Neural Networks

Interpretability Practitioners

Authorship Attribution

Fake News Midterm

Easily Fooling Deep Neural Networks

Multi-Agent Diverse Generative Adversarial Networks

Adversarial Explanations

Topics covered

Popular Clips

Episode Highlights

Neural Network ExplanationsWalt Woods explores advancements in neural network interpretability, focusing on adversarial explanations to enhance image classification decisions. He discusses methods to improve explanation quality and critiques the black box nature of these models.

Neural Network Explanations

Improving Network Robustness

Adversarial ExamplesWalt Woods explores the intriguing world of adversarial examples in neural networks, revealing their potential to mislead image classification systems. He discusses the implications for neural network robustness and the challenges of addressing these vulnerabilities.

Adversarial Examples

Related Episodes