CSE 517a Roadmap

This outline is tentative and may change in the course of the semester.

  1. Structural Risk Minimization
    • Loss Functions and Regularizers
    • Beyond Gradient Descent: Newton’s Method and SGD
  2. Probability Estimation (MLE, MAP, and Bayesian parameter estimation)
  3. Performance Evaluation for ML
  4. Naive Bayes and Generative ML models
  5. Kernels
    • Theory
    • Kernel Machines (Kernel Regression, Kernel SVM)
  6. Gaussian Processes
    • Regression
    • Model Selection
  7. Neural Networks
    • Deep Learning
    • Autoencoders
    • Convolutional NNs
  8. Unsupervised Learning I
    • k-means Clustering
    • GMMs and EM
  9. Unsupervised Learning II
    • PCA and SVD for Dimensionality Reduction
    • Latent Variable Models (PPCA)
  10. Semi-supervised Learning
    • graph-based SSL
    • SSL with SVMs

Optional Topics – time permitting

  1. Graph-based Machine Learning
    • Structured Data
    • Graph Kernels
  2. Active Learning
  3. Sampling and Advanced Mixture Models
    • MCMC / Gibbs sampling
    • Dirichlet Processes, LDA for topic modeling
  4. Large-scale ML
    • parallelizing ML algorithms
    • large-scale linear algebra
    • sparse GPs