Course Details

Course details of EE5180W
Course NoEE5180W
Course TitleIntroduction to Machine Learning
Course Content1. An introduction to machine learning: why and what. A comparison of artificial intelligence, machine learning, and widely adored deep neural networks. 2. The most fundamental problem of electrical engineering: decision making under uncertainty (elaborated with examples from communication and signal processing). Detection and estimation theory & machine learning: similarities and differences.3. Supervised learning (discrete labels): signal detection without the knowledge of path loss and noise distribution, image recognition, etc. Linear classifier, support vector machine and kernel method. Logistic regression. 4. Supervised learning (continuous labels a.k.a. function learning): LTI system and channel estimation. Linear regression, support vector regression.5. A brief tour of neural networks. Why function representation? Why NN? Why deep NN? Some architectures: convolutional neural networks (image processing), recurrent neural networks (communication and control). Training, backpropagation and SGD.6. Unsupervised learning: vector quantization and clustering, k-means algorithm, spectral clustering7. Sparse recovery: applications in signal processing. LASSO, ISTA.8. Low dimensional structure in high dimensional data: PCA9. Graphical model: a statistical model for error correction codes, social networks, etc. Markov random field (MRF), inference on MRF, learning MRF structure from data.10. Reinforcement learning: applications in robotics and wireless scheduling. A brief introduction to Markov decision processes, TD(λ) and Q-learning.
Course Offered this semesterNo
Faculty Name