Machine learning

MACHINE LEARNING CS 229 / STATS 229 INTRODUCTION GOALS

By the end of the course, Be an expert in ML (understand the internals of ML algorithms) Be able to build ML applications (know which algorithms to use when) Be able to start ML research (read research papers) PREREQUISITES Basic computer science principles Big-O notation

Comfortably write non-trivial code in Python/numpy Probability (CS 109, STATS 116 etc.) Random Variables Expectations Distributions Linear Algebra & Multivariate/Matrix Calculus (MATH 51, etc.)

Gradients and Hessians Eigenvalue/vector HONOR CODE Form study groups (strongly encouraged!) Independently write-up homework and code It is an honor code violation to intentionally refer to a previous

year's assignments. COURSE STRUCTURE 3 Homeworks : 60% (3 x 20%) Final exam (take home) : 40% LOGISTICS Course website (calendar, deadlines, notes) - http://cs229.stanford.edu

Piazza Gradescope WHAT IS ML? Term Machine Learning coined by Arthur Samuel in 1959. Samuel Checkers-playing Program Common definition (by Tom Mitchell):

Machine Learning is the study of computer algorithms that improve automatically through experience Subfield of Artificial Intelligence (AI) The hottest subfield - reinvigorated interest in AI due to deep learning! RECENT PROGRESS

Computer Vision / Image Recognition ImageNet Convolutional Neural Networks Autonomous driving Speech Recognition Voice assistants

Language Translation Google Translate Unsupervised Translation Game Playing / Deep Reinforcement Learning ATARI AlphaGo

COURSE PREVIEW Supervised learning [Input -> Output mapping] Classification vs Regression Parametric vs Non-parametric Generative vs Discriminative Probabilistic vs Non-probabilistic Unsupervised learning [Learn interesting structures in the data]

Clusters vs Subspaces Probabilistic vs Non-probabilistic Deep Learning [Learning representations] Our focus: supervised setting Learning Theory Bias-Variance Tradeoff

Generalization and Uniform Convergence Reinforcement Learning [Sequential Decision Making] PREVIEW - SUPERVISED PREVIEW UNSUPERVISED (ICA)

Mixed Separated PREVIEW UNSUPERVISED (ICA - II) Mixed

Separated PREVIEW REINFORCEMENT LEARNING