Distributions and Collections in Machine Learning

Description

Distributions and Collections in Machine Learning is a seminar-style, research course geared towards exploring the advance treatment of collections like sets, distributions, sequences, and graphs in machine learning. A tentative list of topics include: traditional density estimation; nonparametric statistics; recurrent neural networks; autoregressive density estimators; autoencoders; VAEs; GANs; GraphNN architectures, differentiable computing, word embeddings and language models; vision approaches, multimodal data and data-fusion; one/few-shot learning; reinforcement learning; attention models; wavelets and functional PCA; random features/projections; dimensionality reduction; seq2seq and pointer networks; multitask/transfer learning; knowledge-bases; kernel methods/MMD/two sample tests.

The class will be an active, participation based course where students are expected to engage in and lead discussions on topics, as well as design lectures for the course in groups. A large component of the course will be a project, where students are expected to develop novel research that may be submitted to top ML and AI conferences.

Note: students enrolled should have a strong understanding of the basic concepts and methods of machine learning. It is strongly suggested that students have taken either an ML, NLP, computer vision, or other related course.