Explore the fundamental concepts of probability essential for machine learning in this 45-minute video lecture. Delve into the challenges of mathematically rigorous probability theory and discover why negative logarithms of probabilities, known as "surprises," are prevalent in machine learning. Learn how probability behaves like mass, how surprises relate to loss functions, and why they are preferable to densities. Examine the connection between Gaussians, probability, and linear algebra. Access accompanying slides and exercise notebooks for hands-on practice. Gain valuable insights into the Math for Machine Learning series, with timestamps provided for easy navigation through key topics.
Overview
Syllabus
Introduction
Probability is subtle
Overview of takeaways
Probability is like mass
Surprises show up more often in ML
Surprises give rise to loss functions
Surprises are better than densities
Gaussians unite probability and linear algebra
Summary of the Math4ML ideas
Additional resources on Math4ML
Taught by
Weights & Biases
