Course materials
Lectures
Lecture 1: Introduction slides VideoLecture 2: Linear prediction slides Video
Lecture 3: Maximum likelihood slides.pdf Video
Lectures 4 & 5: Regularizers, basis functions and cross-validation slides.pdf Video 1 Video 2
Lecture 6: Optimisation slides.pdf Video
Lecture 7: Logistic regression slides.pdf Video
Lecture 8: Back-propagation and layer-wise design of neural nets slides.pdf Video
Lecture 9: Neural networks and deep learning with Torch slides.pdf Video
Lecture 10: Convolutional neural networks slides.pdf Video
Lecture 11: Max-margin learning and siamese networks slides.pdf Video
Lecture 12: Recurrent neural networks and LSTMs slides.pdf Video
Lecture 13: Hand-writing with recurrent neural networks (Guest speaker: Alex Graves from Google Deepmind)
Lecture 14: Variational autoencoders and image generation (Guest speaker: Karol Gregor from Google Deepmind)
Lecture 15: Reinforcement learning with direct policy search slides.pdf
Lecture 16: Reinforcement learning with action-value functions slides.pdf
Practicals
Please click on Timetables on the right hand side of this page for time and location of the practicals. The instructors are Brendan Shillingford and Marcin Moczulsky.
Practicals will use Torch, a powerful programming framework for deep learning that is very popular at Google and Facebook research.
Practical on week 2: (1) Learning Lua and the tensor library. pdfPractical on week 3: (2) Online and batch linear regression. pdf
Practical on week 4: (3) Logistic regression and optimization. pdf
Practical on week 5: continued previous practical.
Practical on week 6: (4) Feedforward neural networks, and implementing your own layer. pdf
Practical on week 7: (5) Intro to nngraph for graph-shaped modules. pdf
Practical on week 8: (6) Training a LSTM language model. pdf
See the Github repository list for the practicals' code and technical instructions.
Classes
Please click on Timetables on the right hand side of this page for time and location of the classes. The exercises appear below and are due Thursdays at 1pm on the specified week.
Class on Week 3: Problem set. Due 1pm Thursday of Week 2.Class on Week 5: Problem set. Due 1pm Thursday of Week 4.
Class on Week 7: Problem set. Due 1pm Thursday of Week 6.
Class on Week 8: Problem set. Due 1pm Thursday of Week 7.
