Lectures and Handouts

 


              Subjects Covered

  1. 1.Introduction and prerequisites refresher:  Course goals, linear algebra and probability. Lecture 1 - Intro, Linear Algebra.pdf - Lecture 2 - Probability, Statistics, PE.pdf

  2. 2.Representation and interpretation of signals:  Human perception of signals, sampling, quantization, the frequency domain, image and sound representations. Lecture 3 - Time Series.pdf - Lecture 4 - Perception and Fixed Features.pdf

  3. 3.Feature analysis and discovery: Useful fixed transforms (DCT, etc), adaptive transforms (KLT/PCA/EM-PCA/online-PCA), feature extraction from familiar signals (audio, video), eigenfaces. Lecture 5 - PCA.pdf

  4. 4.More feature analysis and dimensionality reduction: Independent Component Analysis (ICA), Non-Negative Matrix Factorization (NMF), Kernel PCA, Manifold embedding methods, random projections. Lecture 6 - ICA-NMF.pdf - Lecture 7 - Non-linear Dim Red.pdf


               Problem Set 1 - Problem Set 1 Solutions.pdf


  1. 5.Detection and classification: Matched filters, template matching, object detection, similarity measures, face detection, speech detection.  Linear classifiers, linear discriminant analysis. Lecture 8 - Detection and Matched Filters.pdf - Lecture 9 - Decision theory and simple classifiers.pdf

  2. 6.Classification: Non-linear classifiers, neural nets, kernels, generative models, non-parametric methods.  Real-world applications of classification models. Lecture 10 - Nonlinear classifiers.pdf - Lecture 11 - Classification bits and pieces.pdf


               Problem Set 2 - Problem Set 2 Solutions.pdf


  1. 7.Clustering interlude. K-means, Gaussian Mixture Models, Expectation-Maximization algorithm. Lecture 12 - K-means, GMMs, EM.pdf

  2. 8.Time series and dynamical models: Classification and similarity, time warping models, Markov models. Lecture 13 - Time series, HMMs, DTW.pdf

  3. 9.Mixed signals: Array processing, beamforming, independent component analysis, MIMO/SIMO models, underconstrained separation, spectral factorizations. Lecture 14 - Arrays and source separation.pdf - Lecture 15 - Underconstrained signal separation.pdf

  4. 10.Matrix factorizations and bag-of-features models: Non-negative Matrix Factorization and Probabilistic Latent Semantic Decompositions, bag models, Convolutive decompositions. Lecture 16 - Matrix Factorizations and beyond.pdf

  5. 11.Missing data techniques and tracking.  Lecture 17 - Missing data and tracking.pdf

  6. 12.Invited Talk: Computer Vision

  7. 13.Machine listening and music information retrieval. Lecture 18 - Machine Listening and MIR.pdf

  8. 14.Invited Talk: Mark Hasegawa-Johnson Semi-Supervised Learning for Speech and Audio Processing.pdf

  9. 15.Invited Talk: Compressive Sensing - Sparsity, Randomness and Compressive Sensing.pdf

  10. 16.Final project presentations