| Week 1 : straw | Week 2 : distributions | Week 3 : feed-forward NNs |
|---|---|---|
| introduction | Gaussian, multinomial, softmax, sigmoid | polynomials (Bishop) |
| k-NN, Perceptron, PCA, k-means, the curse | losses and likelihoods | Backprop |
| the curse | (no lecture) | basics lab |
| Week 4 : more FFNN | Week 5 : deep papers | Week 6 : recurrent NNs |
| convnets; varieties of SGD | papers 2 | basic recurrent NNs |
| weight decay; dropout | papers 3 | adv. recurrent NNs |
| papers 1 | papers 4 | Q & A |
| (break) | ||
| Week 7 : Bayes | Week 8 : directed PGMs | Week 9 : inference over time |
| language of inference | belief nets | eg: HMMs |
| Bayesian predictions | sum-product, max-sum | Kalman filters, particle filters |
| Gaussian processes | eg: mixtures of Gaussians | Q & A |
| Week 10 : undirected PGMs | Week 11 : deep generative models | Week 12 : the next big thing |
| Boltzmann machines | papers 4 | discussion |
| Restricted BMs | papers 5 | 3 crazy ideas |
| deep generative | papers 6 | wrap up |
- 5 for each of two presentations
- 10 for each of three assignments
- 20 for a project (and short interview on it)
- 40 for a 2 hour exam
Pull requests Add some notes of what you lecture up here, scribbles/notes of related problems, add links to resources, ... ?
Raise an issue. Ask questions, clarify course matrial, ... ?? Just treat it like a forum.
These will be the notes and resources the exams and assignments are built from. It should (hopefully) serve as good study material.