Content

Current issues, methods and datasets of unsupervised learning in image and text processing, including LSTMs, transformers, generative models.  All lecture material such as slides, recordings, as well as the exercises are available in the lecture moodle.

Programming exercises will be implemented in python and pytorch.

Schedule

• Start: Monday 10.10.2022, 10-12am
• Lecture: Monday  from 10.10.2022, 10-12pm
• Exercise: Monday, from 10.10.2022, 12am-2pm

Qualification Goals

• Deep understanding of current methods of unsupervised learning of image and text representations, self-supervised learning, representation learning, generative models.
• Understand, apply and evaluate current approaches.
• Understanding the technical underpinnings of unsupervised learning methods.
• Evaluate and discuss new learning problems and unsupervised and self-supervised methods.

Literature

• Ian Goodfellow and Yoshua Bengio and Aaron Courville, Deep Learning, An MIT press book, 2016.
• Attention and Augmented Recurrent Neural Networks, Chris Olah and Shan Carter. Distill, 2016
• Generating Sequence with Recurrent Neural Networks, A. Graves, ArXiV