INT 86WK: MACHINE LEARNING

University of California, Santa Barbara

The ability of recently engineered machine learning algorithms and natural systems to perform inference and generalize well from limited finite data o bservations poses interesting challenges and open problems. This seminar wi ll discuss both practical algorithms for applications and related rigorous mathematical theory. Examples include the approximation and generative abi lities of deep learning with recent types of neural networks, formulations and training of unsupervised methods such as transformers, diffusion-models , autoencoders, and non-neural network approaches such as support vector ma chines, kernel methods, and probabilistic methods.