Towards Fair Deep Anomaly Detection

Explore a groundbreaking conference talk on fair deep anomaly detection presented at FAccT 2021. Delve into the research by H. Zhang and I. Davidson as they address the challenges of bias in machine learning algorithms. Learn about Deep SVDD, anonymity detection, and the proposed framework utilizing adversarial learning to minimize unfairness. Examine the experimental results and gain insights into the potential applications of this innovative approach for creating more equitable AI systems.

Syllabus

Introduction
Deep SVDD
Anonymity Detection
Outline
The 80 Rule
Proposed Framework
adversarial learning
minimizing adversarial loss
pipeline
extensions
Experiments
Results
Experimental Results
Summary