CSE329

---

Machine Learning

3 Credit Hour Course

Intended For Level 3 Term 2 Students

Prerequisite: CSE301, CSE317

Developing machine learning systems: problem formulation, data collection, manipulation and preprocessing, exploratory data analysis and visualization; Deep learning: linear regression as a neural network, simple feedforward networks, forward propagation, backward propagation, computation graphs, numerical stability and initialization; Optimization: batch and stochastic gradient descent (SGD); Convolutional neural networks (CNN): convolution, padding, stride, pooling, modern CNNs (AlexNet, VGG, GoogLeNet, ResNet), batch normalization; Recurrent neural networks (RNN): language modeling with RNNs, modern RNNs, long short-term memory (LSTM), gated recurrent units (GRU), recursive neural networks, sequence-to-sequence (seq2seq) models; Generalization in deep learning: designing neural network architectures, weight decay, dropout; Probabilistic modeling and reasoning: Bayesian networks, exact inference, variable elimination algorithm, approximate inference, direct sampling methods, inference by Markov chain simulation, Gibb’s sampling; Probabilistic reasoning over time: inference in temporal models, hidden Markov model, Kalman filters; Learning probabilistic models: learning with complete data, Bayesian learning, naive Bayes models, generative and discriminative models, generalized linear model; Learning with hidden variables: expectation–maximization (EM) algorithm, mixture models, learning mixtures of Gaussians, K-means clustering, learning hidden Markov models; Dimensionality reduction: principal component analysis (PCA); Recommender systems: collaborative filtering using matrix factorization.