This repository contains all colabs and links to material for the GSERM Summerschool (Winterschool) of the University of St. Gallen (HSG), Switzerland for 10,860,1.00 - Introduction to Applied Deep Learning.
This course is using TensorFlow 2.1.0 and Google Colab.
| # | Lecture | Lab |
|---|---|---|
| 1 | Introduction to Deep Learning Lecture Slides Topics: Perceptron, Feed Forward Neural Networks, Deep Fully Connected Neural Networks, Activation and Loss Function, Optimizers, Overfitting, Dropout and other Regularization |
Colab L1 - Deep Learning Intro with mnist |
| 2 | Deep Computer Vision Lecture Slides Topics: Convolution, Subsampling, GPUs, Batchnorm, TensorBoard, Spatial Invariance, CNN Applications, Transfer Learning, ImageNet, TF Hub, TF Datasets |
Colab L2 - Convolutional Neural Networks Colab L3 - Transfer Learning |
| 3 | Deep Sequence Modeling Lecture Slides Topics 1: Modern Convolutional Neural Networks, Inception, SqueezeNet, Keras Functional API, Data Processing with tf Datasets, TPUs, TF Production with TFX Topics 2: Sequential Data, RNN, LSTM, GRU |
Colab L3 - Data Processing Colab L3 - Recurrent Neural Networks |
| 4 | Deep Natural Language Processing Lecture Slides (Guest Lecture by ejoebstl) Topics: Language in Deep Learning, Word2Vec, Seq2Seq, Neural Machine Translation, Attention, Transformer, BERT |
Colab L4 - Word Embeddings with Word2Vec Colab L4 - Seq 2 Seq / Attention for translations Colab L4 - Bert and Keras with transfer learning |
| 5 | Data Bias and Explainable Deep Learning Lecture Slides Topics: Recipe for Training Neural Networks, Data Bias, Deep Learning in Academic Context and for Academic Publications, Explainable Artificial Intelligence (XAI) |
Colab L5 - Interpreting Vision Models with tf-explain. Colab L5 - Visualize CNN Layer Outputs and Filters with Keras Functional API. |
Book
If there would be a book or resources I'd want share with you they are the following: Deep Learning Book. You can find the full PDF in their GH repo. This is a very comprehensive book that is up to date and good to read with a valuable Bibliography.
Courses
MIT's 6.S191: Introduction to Deep Learning is a good course covering also Reinforcment Learning and GANs. Stanford's CS231n is a good place to dive deeper into the mechanics of computer vision.
Other Materials
The (IMO) didactically best material on deep learning is Google, Tensorflow and deep learning without a PhD series, by Martin Görner. The official Tensorflow Resources are well done and helpful. There are a big number of examples, tutorials and other resources available. Also helpful is their page on additional learning resources.
Main Sources and Credits:
Google, Tensorflow and deep learning without a PhD series, Martin Gorner
MIT 6.S191: Introduction to Deep Learning
Tensorflow API Documentation and Resources
How to use this material?
All course materials are copyrighted by their respective author and owner. If you are an instructor and would like to use any materials from this course (slides, labs, code) for educational purposes, you must reference the original source and this repository.
This repository is open sourced under Apache-2.0 License but the license of the original author might apply.