Pattern Recognition
| Code | 3.3.3208.9 |
|---|---|
| Semester | 9th |
| Flow | S - Signals, Automatic Control and Robotics |
| Category | Obligatory by selection |
| Credits | 4 |
| Class Hours - Lab Hours | 3 - 1 |
| Lecturers | Alexandros Potamianos |
| Links | Helios, Course's Website |
| Web Platform |
Class 1:
Webex
|
Description
Statistical pattern recognition with application to recognition and tracking of audio, visual or audio-visual objects and events, as well as, other spatiotemporal sensory or symbolic data. The course focuses on the processing of audio and speech signals, natural language processing, computer vision, and multimodal signal processing, with emphasis on estimation & learning theory of sequences. In detail: Machine learning review. Advanced parameter estimation: maximum a posteriori, Bayesian learning, expectation maximization, unsupervised learning. Advanced optimization theory with application to SVMs. Advanced feature selection and feature transformation methods: Independent Component Analysis (ICA), Non-negative Matrix Factorization (NMF), Latent Semantic Analysis (LSA). Probabilistic Graphical models: Markov models, Hidden Markov Models (HMMs), Conditional Random Fields (CRFs), Markov Random Fields (MRFs), inference on graphical models. Review of deep learning (DNNs, CNNs). Neural models for sequences: RNNs, gating (LSTMs, GRUs), encoder-decoders, attention, transformers, auto-encoders, deep generative models, GANs, graph NNs. Structured probabilistic models, Monte Carlo methods, approximate inference, partition function estimation. Neuro-symbolic NNs, memory models, cognitive architectures. Advanced reinforcement learning: value-based, policy gradient, actor-critic. Geometry of meaning: intro to metric and topological spaces, Riemannian manifolds, Poincare embeddings.