Navigation

19D051DU - Deep Learning

Course specification
Course title Deep Learning
Acronym 19D051DU
Study programme
Module
Type of study
Lecturer (for classes)
Lecturer/Associate (for practice)
    Lecturer/Associate (for OTC)
      ESPB 9.0 Status elective
      Condition
      The goal Introduction to modern Artificial Intelligence techniques, based mainly on Machine Learning. Understanding of prerequisites, potentials, limitations and methods of applying these techniques to problems of processing images, videos, sequential signals, and natural language, control of agents, generation of images and text, visualization of high-dimensional data.
      The outcome Students understand the basic and advanced concepts of studied techniques, they can implement them using standard libraries, train the models and analyze their performance, introduce modifications with the goal of adapting existing methods to specific domains, or as part of scientific research.
      Contents
      URL to the subject page https://www.etf.bg.edu.rs/fis/karton_predmeta/19D051DU-2023
      Contents of lectures Models for working with images: VGG, ResNet, ViT. Object detection and segmentation: YOLO, Faster R-CNN, Mask-RCNN, UNet. Object tracking: DeepSORT. Sequential data, attention: LSTM/GRU, Transformer. Graph Neural Networks. Reinforcement Learning: DDPG, DQN, PPO. Self-supervised learning: BERT, GPT, SimCLR. Generative models: GAN, VAE, diffusion models. Dimensionality reduction: UMAP, t-SNE.
      Contents of exercises Independent implementation, training and evaluation of models using the Python programming language and standard libraries (PyTorch, TensorFlow, Keras).
      Literature
      1. Goodfellow, Ian, Yoshua Bengio, and Aaron Courville. Deep learning. MIT press, 2016. (Original title)
      2. Sutton, Richard S., and Andrew G. Barto. Reinforcement learning: An introduction. MIT press, 2018. (Original title)
      3. Murphy, Kevin P. Probabilistic machine learning: an introduction. MIT press, 2022. (Original title)
      4. Bronstein, Michael M., et al. Geometric deep learning: Grids, groups, graphs, geodesics, and gauges. arXiv preprint arXiv:2104.13478 (2021). (Original title)
      5. Zhang, Aston, et al. Dive into deep learning. Cambridge University Press, 2023. (Original title)
      Number of hours per week during the semester/trimester/year
      Lectures Exercises OTC Study and Research Other classes
      8
      Methods of teaching Lectures. Independent research of the suggested literature and additional seleted scientific papers, with consultations.
      Knowledge score (maximum points 100)
      Pre obligations Points Final exam Points
      Activites during lectures Test paper
      Practical lessons Oral examination 30
      Projects 70
      Colloquia
      Seminars