Machine Learning Biases

This lecture covers the basics of machine learning, including supervised, unsupervised, and reinforcement learning. It delves into the challenges of real-world machine learning deployment, such as adversarial attacks and privacy concerns. The instructor explains how machine learning models are trained, tested, and deployed, using examples like hate speech detection and patient hospitalization prediction. The lecture also explores the ubiquity of machine learning in various fields and the implications for security and privacy. Adversarial conditions in machine learning are discussed, focusing on confidentiality, integrity, and availability threats. Different types of attacks, such as black-box, grey-box, and white-box attacks, are explained, along with strategies to prevent model stealing.