lsml22

Notebooks for the 2022 course on Large Scale Machine Learning at Mines ParisTech.

Day 1: Introduction to large scale ML

• If you are familiar with scikit-learn (for example, Mines students who had Data Science in their first year or took "Apprentissage artificiel" previously), work on notebook 1_sklearn_at_scale.ipynb.
• If you are not familiar with scikit-learn, you can start with this notebook to get a hang of things.

Day 2: Deep learning

• If you have never trained convolutional neural networks on keras, start with this notebook to train a LeNet Deep Convolutional Network on MNIST.
• Practice transfer-learning using a standard ConvNet pre-trained on ImageNet with this notebook
• Practice unsupervised deep learning with auto-encoders and GAN with this notebook Beginners should work on TP1 (LeNet on MNIST), and then at least begin TP3 (Deep Generative Models) Students who have already practised with Deep ConvNets should work essentially on TP3. TP2 may be useful only for those who have never practised Transfer Learning.

Day 3: Stochastic gradient descent

Work on notebook 3_stochastic_gradient_descent.ipynb.

Day 4: Deep reinforcement learing

Instructions inside 4_deep_reinforcement_learning.pdf. The notebook is here.

Setup

To run the notebooks, you will need Python, Jupyter (either JupyterLab or Jupyter Notebook), and number of Python librairies. The easiest way to install of this is to use conda and set up an environment specific to this course using the file package_list.yml. To this end, you can either:

• if you prefer graphical user interfaces: (1) install Anaconda and (2) follow the instructions under "Importing an environment" of the tutorial to import the environment in package_list.yml;
• if you prefer the command line: (1) install conda and (2) use the following instructions in the command line:

   conda env create -f package_list.yml -n lsml
   conda activate lsml

Course materials

You can find course materials here: https://cloud.mines-paristech.fr/index.php/s/eV67rL3ySnN4JIn