Study program: F3 Computer Science

Degree: Master

Type of module: lectures, practical work

Lecturer: Assoc. Prof., PhD Zulfiia Imanhulova

Language: English

Credit Points: 5 ECTS

Description:
The discipline is dedicated to deep learning methods, a new generation of neural network machine learning methods that have caused a rapid surge in development in a number of application areas. The discipline is primarily aimed at developing students' skills in solving applied problems using deep neural networks.

Content module 1. Introduction to deep learning.
1.1. Supervised learning in neural networks (NN).
1.2. Gradient descent for NN.
1.3. Forward and backward propagation. Parameters and hyperparameters in NN.
Content module 2. Improving deep NN: tuning hyperparameters, regularization and optimization.
2.1. Bias and variance in NNs.
2.2. Methods of regularization of NN parameters.
2.3. Training on mini batches.
2.4. Algorithms for optimizing NN parameters.
Content module 3. Convolutional NN.
3.1. Computer vision.
3.2. Structure of convolutional NN. Different models of convolutional NN.
3.3 Object detection task.