Topic outline

    • Contents: Machine learning with deep neural networks. Programming using PyTorch.

    After the course, the student understands the basic principles of deep learning: fully-connected, convolutional and recurrent neural networks; stochastic gradient descent and backpropagation; means to prevent overfitting. The student understands how to do supervised learning (classification and regression) and unsupervised learning with neural networks. The student knows modern neural architectures used for image classification, time series modeling and natural language processing. The student has experience on training deep learning models in PyTorch.

    • Assessment: Exercises, a project work and an exam.

    • Prerequisites: Basics of machine learning, basics of probability and statistics, good level of programming in Python. Recommended: matrix algebra.
    Basic terms of machine learning:
      • supervised and unsupervised learning
      • overfitting and underfitting
      • regularization
    Basic terms of probability theory:
      • sum, product rule, Bayes' rule
      • expectation, mean, variance, median
    • Course contents:
      • Introduction and history of deep learning
      • Optimization for training deep models
      • Regularization for deep learning
      • Convolutional networks
      • Recurrent neural networks
      • Unsupervised learning with deep autoencoders and generative adversarial networks