Fork of Lempitsky DL for HSE master students
Lecture and seminar materials for each week is in ./week* folders
- Create cloud jupyter with repo https://beta.mybinder.org/v2/gh/yandexdataschool/Practical_RL/fall17
- Any technical issues, ideas, bugs in course materials, contribution ideas - add an issue
- Bookmark repo https://github.com/yandexdataschool/practical_deeplearning
- Join telegram chat https://t.me/dl_hse_fall17
- Enroll to anytak TBA
- Join piazza https://piazza.com/cs_hse/fall2017/dl101/home with access code dl101
- 06.09 - Course started
- week0 Recap
- Lecture: Linear models, stochastic optimization, regularization
- Seminar: Linear classification, sgd, modifications
- HW due: 28.09.16, 23.59.
- Please get bleeding edge theano+lasagne installed for the next seminar.
- week1 Getting deeper
- Lecture: Neural networks 101
- Seminar: theano, symbolic graphs and basic neural networks
- HW due: 3.10.16 23.59
- week2 Deep learning for computer vision 101
- Lecture: Convolutional neural networks
- Seminar: lasagne and CIFAR
- HW due: 9.10.16 23.59 on first submission.
- week3 Deep learning for natural language processing 101
- Lecture: NLP problems and applications, bag of words, word embeddings, word2vec, text convolution.
- Seminar: Text convolutions for Avito content filtering task
- HW due: 16.10.16 23.59 on first submission.
- week4 Recurrent neural networks for sequences
- Lecture: Simple RNN. Why BPTT isn't worth 4 letters. GRU/LSTM. Language modelling. Optimized softmax. Time series applications.
- Seminar: Generating laws for pitiful humans with mighty RNNs.
- HW due: 28.10.16 23.59 on first submission.
- week5 Recurrent neural networks II
- Lecture: Batchnorm and dropout for RNN; Seq2seq: machine translation, conversation models, speech recognition and more. Attention. Long term memory architectures.
- Seminar: a toy machine translation task
- to be anounced
- [Skip week]
- week6 Fine-tuning with neural networks
- Lecture: Large CV datasets, model zoo, reusing pre-trained networks, fine-tuning, "knowledge transfer", soft-targets
- Seminar: Cats Vs Dogs Vs Very Deep Networks
- HW due 17.11.16 23.59
- week7 Advanced computer vision
- Lecture: Representations within convnets, fully-convolutional networks, bounding box regression, maxout, etc.
- Seminar: Image captioning by Arseniy Ashukha
- HW due 24.11.16 23.59
- week8: Generative models for computer vision
- Lecture: Autoencoders, Generative Adversarial Networks
- Seminar: Art Style Transfer with deep learning (Dmitry Ulyanov)
- HW due 4.12.16 23.59
- week8: Deep learning for sound processing
- Lecture: case study: music recommendation with deep learning
- Seminar: Music clustering & content-based recommentation with convolutional nets
- HW due 11.12.16 23.59
- week9: Basic reinforcement learning
- Lecture: Introduction to reinforcement learning
- Seminar: one algorithm to navigate in a maze, play pacman and control robots.
- HW due 11.12.16 23.59
- week10: Deep reinforcement learning
- Lecture: approximate reinforcement learning with deep neural networks (problems and solutions)
- Seminar: Playing Atari/Doom with deep reinforcement learning
- HW due 18.12.16 23.59 first submission
- week12: Bayesian deep learning
- Lecture: Basics of bayesian approach to probabilities
- Bonus lecture: Variational autoencoders (Mikhail Khalman)
- HW due 26.12.16 16.00 hard
- One rule to rule them all
- Project rules
- Project examples
- Reducing lateness penalty
- Feedback form (anonymous)
Course materials and teaching performed by
- Fedor Ratnikov - lectures, seminars, hw checkups
- Oleg Vasilev - seminars, hw checkups, technical issue resolution
- Arseniy Ashukha - image captioning, sound processing, week7&9 lectures
- Dmitry Ulyanov - generative models, week8 lecture, week12 homework assignment
- Mikhail Khalman - variational autoencoders, lecture 12
- Vadim Lebedev - week0 & week6 homeworks
