A propos

This repository contains all the documents and links of the Fidle Training .
Fidle (for Formation Introduction au Deep Learning) is a 2-day training session
co-organized by the Formation Permanente CNRS and the SARI and DEVLOG networks.

The objectives of this training are :

• Understanding the bases of Deep Learning neural networks
• Develop a first experience through simple and representative examples
• Understanding Tensorflow/Keras and Jupyter lab technologies
• Apprehend the academic computing environments Tier-2 or Tier-1 with powerfull GPU

For more information, you can contact us at :
Current Version : 2.0.1

Course materials

Course slides The course in pdf format (12 Mo)	Notebooks     Get a Zip or clone this repository      (10 Mo)	Datasets All the needed datasets (1.2 Go)

Have a look about How to get and install these notebooks and datasets.

Jupyter notebooks

Linear and logistic regression

• LINR1 - Linear regression with direct resolution
  Low-level implementation, using numpy, of a direct resolution for a linear regression
• GRAD1 - Linear regression with gradient descent
  Low level implementation of a solution by gradient descent. Basic and stochastic approach.
• POLR1 - Complexity Syndrome
  Illustration of the problem of complexity with the polynomial regression
• LOGR1 - Logistic regression
  Simple example of logistic regression with a sklearn solution

Perceptron Model 1957

• PER57 - Perceptron Model 1957
  Example of use of a Perceptron, with sklearn and IRIS dataset of 1936 !

Basic regression using DNN

• BHPD1 - Regression with a Dense Network (DNN)
  Simple example of a regression with the dataset Boston Housing Prices Dataset (BHPD)
• BHPD2 - Regression with a Dense Network (DNN) - Advanced code
  A more advanced implementation of the precedent example

Basic classification using a DNN

• MNIST1 - Simple classification with DNN
  An example of classification using a dense neural network for the famous MNIST dataset

Images classification with Convolutional Neural Networks (CNN)

• GTSRB1 - Dataset analysis and preparation
  Episode 1 : Analysis of the GTSRB dataset and creation of an enhanced dataset
• GTSRB2 - First convolutions
  Episode 2 : First convolutions and first classification of our traffic signs
• GTSRB3 - Training monitoring
  Episode 3 : Monitoring, analysis and check points during a training session
• GTSRB4 - Data augmentation
  Episode 4 : Adding data by data augmentation when we lack it, to improve our results
• GTSRB5 - Full convolutions
  Episode 5 : A lot of models, a lot of datasets and a lot of results.
• GTSRB6 - Full convolutions as a batch
  Episode 6 : To compute bigger, use your notebook in batch mode
• GTSRB7 - Batch reports
  Episode 7 : Displaying our jobs report, and the winner is...
• GTSRB10 - OAR batch script submission
  Bash script for an OAR batch submission of an ipython code
• GTSRB11 - SLURM batch script
  Bash script for a Slurm batch submission of an ipython code

Sentiment analysis with word embedding

• IMDB1 - Sentiment alalysis with text embedding
  A very classical example of word embedding with a dataset from Internet Movie Database (IMDB)
• IMDB2 - Reload and reuse a saved model
  Retrieving a saved model to perform a sentiment analysis (movie review)
• IMDB3 - Sentiment analysis with a LSTM network
  Still the same problem, but with a network combining embedding and LSTM

Time series with Recurrent Neural Network (RNN)

• SYNOP1 - Preparation of data
  Episode 1 : Data analysis and preparation of a meteorological dataset (SYNOP)
• SYNOP2 - First predictions at 3h
  Episode 2 : Learning session and weather prediction attempt at 3h
• SYNOP3 - 12h predictions
  Episode 3: Attempt to predict in a more longer term

Unsupervised learning with an autoencoder neural network (AE)

• AE1 - Building and training an AE denoiser model
  Episode 1 : After construction, the model is trained with noisy data from the MNIST dataset.
• AE2 - Exploring our denoiser model
  Episode 2 : Using the previously trained autoencoder to denoise data

Generative network with Variational Autoencoder (VAE)

• VAE1 - First VAE, with a small dataset (MNIST)
  Construction and training of a VAE with a latent space of small dimension.
• VAE2 - Analysis of the associated latent space
  Visualization and analysis of the VAE's latent space
• VAE5 - Another game play : About the CelebA dataset
  Episode 1 : Presentation of the CelebA dataset and problems related to its size
• VAE6 - Generation of a clustered dataset
  Episode 2 : Analysis of the CelebA dataset and creation of an clustered and usable dataset
• VAE7 - Checking the clustered dataset
  Episode : 3 Clustered dataset verification and testing of our datagenerator
• VAE8 - Training session for our VAE
  Episode 4 : Training with our clustered datasets in notebook or batch mode
• VAE9 - Data generation from latent space
  Episode 5 : Exploring latent space to generate new data
• VAE10 - SLURM batch script
  Bash script for SLURM batch submission of VAE8 notebooks

Miscellaneous

• ACTF1 - Activation functions
  Some activation functions, with their derivatives.
• NP1 - A short introduction to Numpy
  Numpy is an essential tool for the Scientific Python.
• TSB1 - Tensorboard with/from Jupyter
  4 ways to use Tensorboard from the Jupyter environment

Installation

A procedure for configuring and starting Jupyter is available in the Wiki.

Licence

[en] Attribution - NonCommercial - ShareAlike 4.0 International (CC BY-NC-SA 4.0)
[Fr] Attribution - Pas d’Utilisation Commerciale - Partage dans les Mêmes Conditions 4.0 International
See License.
See Disclaimer.