Main Textbooks

Slides & Syllabus

• Fundamental building blocks (Basics, MLP, CNN) 

• Recurrent Neural networks (RNN, LSTM, GRU)

• Modern CNNs (Inception, ResNet, MobileNet)

• Optimization (Gradient descent family, Adam, Learning rate scheduler)

• Attention mechanism & Transformers (Seq2Seq,  Attention, Transformers, BERT, GPT, BART)

[recommended videos 1 & 2 for Transformers]

(the corresponding ppsx file for playing gifs & videos in the slides)

• Generative networks (GAN, VAE)

[recommended video for vae]

Codes & Notebooks

Basics:

• Introduction to Pytorch

• Basic Linear Regression using Pytorch

• Basic Logistic Regression using Pytorch

MLP:

• MNIST classification using MLP

CNN:

• MNIST classification using LeNet

RNN:

• Vanilla RNN basics in pytorch

• Name classification using Vanilla RNN [name data]

• Sentiment analysis using LSTM

Optimization:

• Understanding contour maps

• Illustration of the family of Gradient descent approaches in pytorch

• Analysis of optimization path of gradient descent based approaches in numpy

Seq2Seq:

• French to English translation using seq2seq

Transformers:

• Linguistic Acceptability Analysis using BERT

Homeworks 

HW1 (deadline: 19 April)

HW2 (deadline: 28 April)

HW3 (deadline: 1 June)

HW4 [data] (deadline: 20 June)

You should provide a single PDF file containing your report and results for all questions (including programming questions). Do not include code in this report. For each programming question, you should include one notebook file with your code and an explanation of the code. Combine all the files into a single RAR file and send it to DL2024.kntu@gmail.com

MIDTERM: 20th April, 9:00 am

online class link: https://meet.google.com/yht-bnie-hgy