SysEn-5888

A collection tutorial materials for the "SYSEN 5888/6888: Deep Learning" course at Cornell University.

Tutorials:

T1: Environment setup, Jupyter overview, and Python Basics

T1: Slides

An environment setup is provided here for all following tutorials including the installation of Python 3.8 and Jupyter through Anaconda distribution. An overview of the Python programming language is also included.

T2: NumPy, Pandas, Visualizations

A tutorial of NumPy for array operations and different packages for visualization.

T3: Keras and Tensorflow Introduction

A tutorial of the installation and basics of Tensorflow and Keras. Development in the course is mainly built on Tensorflow==2.5.0

T4: Basics of Vision CNN.

A tutorial of the basics of Convolutional Neural Network including operations and MNIST example

T5: CNN in Practice

A tutorial of more complex Convolutional Neural Networks and relevant techniques and algorithms.

T6: Transfer learning

A tutorial of transfer learning including fine-tuning and Tensorflow Hub

T7: Object detection and Segmentation

A tutorial for You Only Look Once V3 (YOLO3v) for object detection and overview of image segmentation

T8: Text and Sequences Working with Data

An introduction to word embedding techniques including Text Vactorization, Skip-gram and Negative Sampling, Bag of Words, and Word2Vec

T9: Text and sequences RNN and CNN

A text classification tutorial for training a recurrent and a convolutional neural networks on the IMDB large movie review dataset for sentiment analysis

T10: Transformers Implementation

A transformer Implementation tutorial training a sequence to sequence (seq2seq) model for Spanish to English translation

T11: Generative Modeling VAE

A tutorial for generative modeling with autoencoders with a basic convolutional autoencoder and variational autoencoder

T12: Generative Modeling GAN

A tutorial for generative modeling with generative adversarial networks for generating handwritten digits, translating image-to-image with a conditional GAN, translating unpaired Image-to-Image using Cycle-GAN

T13: Deep Reinforcement Learning

A Deep Reinforcement Learning tutorial with two examples on: playing CartPole with the actor-critic method, and Deep Deterministic Policy Gradient (DDPG) for the classic Inverted Pendulum control problem

T14: GNN and Best Practices

A tutorial of Graph Neural Networks with Spektral and some best practices in Deep Learning including Keras functional API, monitoring deep-learning models using Keras callbacks and TensorBoard, hyperparameter optimization, model ensembling, multi-GPU and distributed training, and training Keras models with TensorFlow Cloud.