TODO.md

Models

✔️ K Nearest Neighbor

• Note
• Write code
• Debug using IRIS
• More test (optional)

✔️ K Means

• Note
• Write Code
• Debug using WINE
• Implement Soft K-Means (optional)
• Debug using datasets (later)

Try use the precision matrix, i.e. $(a_{ij})$ in which each row is the result of clustering prediction and each column for the actual label. Also, check MNIST 2d visualization

Bishop: life time good book to read

Read later: link

✔️ MLP: fully-connected feed-forward NN

• Note
• Simple MLP: no debug!
• MLP in sklearn
• MLP in pytorch not necessary

✔️ CNN: convolution neural net

• Note See study note in 19W notebook
• Simple Code
• Tutorial by @Puneet Grover Write your own after reading so much stuffs!

RNN: recurrent neural net

• Note
• vanilla RNN with PyTorch official tutorial
• word-level RNN generator with LSTM cells

---

Talks

Reinforce Learning

• Prof. Precup: day 2
• Prof. Precup: day 3
• TBA

Monte Carlo Tree Search

• TBA

Natural Language Processing

• TBA

---

Misc

Click me for github markdown emoji

math in matplotlib

Atom Hydrogen Debug

A Recipe for Training Neural Net