Objectives and outcomes
Introduction to the methodology of problem-solving using machine learning techniques, and training for
independent application of machine learning algorithms. The knowledge of important machine learning
algorithms and methods of their evaluation. Understanding the matrix setting of the problem, as well as
the modular approach to learning layered architectures. Knowledge of optimisation and regularisation
procedures. Understanding linear and logistic regression algorithms. Knowledge of training
methodology for multi-layer neural networks. Learning about ensemble methods and techniques of
dimensionality reduction. Understanding the basic principles of unsupervised and reinforcement learning,
convolutional and recurrent neural networks.
Lectures
Introduction, main steps of machine learning algorithms, ways of model evaluation. Linear regression.
Regularisation and optimisation techniques. Logistic regression. A modular approach to
multi-layer neural networks. Neural networks. SVM algorithms (support vector machines). Dimension
reduction. Decision trees, ensemble algorithms, RDF (random decision forests). Unsupervised learning algorithms. Basics of reinforcement learning. Fundamentals of convolutional and recurrent neural networks. Introduction to deep learning.
Practical classes
Selection and practical design of models covered in lectures. Scikit-learn, TensorFlow. Linear and
logistic regression. Instance-based methods. Language processing, Naive Bayes. Neural networks.
Applications to computer vision, OpenCV. Clustering, dimension reduction. SVM. RDF, boosting
algorithms, XGBoost. Applications to regression, classification and clustering problems.