The course will explore in depth several important classes of algorithms in modern machine learning, and cover applications of each algorithm in real-world settings.
We will cover the following topics.
Methods for linear and non-linear dimensionality reduction. Shallow embeddings and matrix factorization. Random-walk embeddings and their connection with spectral methods. Locality-preserving projections.
Machine learning on relational data. Using machine learning to reason with relational data, especially in the context of knowledge graph embeddings. Network representation learning methods. Graph convolutional networks. End-to-end learning on relational data using low-dimensional embeddings.
Distant supervision, multi-label and multi-target learning. Weak supervision. Data programming. One-shot and few-shot learning. Noise-contrastive estimation (NCE) and optimization.
Introduction to reinforcement learning.
Current research topics.
