Research: Deep learning course
The course on Deepl learning is a part of Computer and information science, master’s study programme and it will be held at the University of Ljubljana, The lectures start at the begging of summer semester (3 lectures per week). The lecturer is Prof. Danijel Skočaj. the course is opened to SMASH fellows and it is held in English.
Below you can see the content of the course :
Introduction to deep learning. Historical perspective. Applications of deep learning.
Training deep neural networks. Feedforward neural networks. Backpropagation. Activation and loss functions. Regularization, initialization, normalization. Learning optimisers.
Convolutional Neural Networks. CNN layers. CNN architectures. Practical methodology. Explainability of CNNs. Adversarial images.
Computer vision. Image classification, object detection, instance segmentation, semantic segmentation, panoptic segmentation. Applications in computer vision.
Recurrent Neural Networks. Dealing with sequential data. Backpropagation through time. RNN. LSTM, GRU. Attention.
Transformers and natural language processing. Transformer architecture. Self-attention. Multi-head attetion. NLP tasks. BERT. BERT extensions. GPT. ChatGPT. Applications in NLP.
Transformers in computer vision. Vision transformer and extensions. Applicatins in computer vision.
Generative methods. Autoencoders, Variational autoencoders, GAN, Normalising flows, Diffusion models, GPT4. Applications in computer vision and NLP.
Advanced concepts. Self-supervised learning. Active, weakly and semi-supervised learning. Deep Reinforcement Learning. Deep learning on graphs.