Slides for the course: [Epita] Deep Neural Networks Course:

• Introduction
• Back Propagation
• Convolutional Neural Networks
• Detection / Segmentation
• Generative Models
• Transformers and MLP-Mixers
• Videos
• DALL-E 2
• Overview
• 3D Scenes

NERF tutorial: gitlab page

These slides are based on another course by Olivier Grisel and Charles Ollion.

Notebooks for the course: [Epita] Deep Neural Networks Course: {{< highlight html >}} wget https://deepcourse-epita.netlify.app/notebooks/Backpropagation_Deepcourse.ipynb wget https://deepcourse-epita.netlify.app/notebooks/Autograd_Deepcourse.ipynb wget https://deepcourse-epita.netlify.app/notebooks/Convolutions_Deepcourse.ipynb wget https://deepcourse-epita.netlify.app/notebooks/CNN_Deepcourse.ipynb wget https://deepcourse-epita.netlify.app/notebooks/Transfer_Deepcourse.ipynb wget https://deepcourse-epita.netlify.app/notebooks/Localization_Deepcourse.ipynb wget https://deepcourse-epita.netlify.app/notebooks/Segmentation_Deepcourse.ipynb wget https://deepcourse-epita.netlify.app/notebooks/Metric_Learning_Deepcourse.ipynb wget https://deepcourse-epita.netlify.app/notebooks/Domain_Adaptation_Deepcourse.ipynb wget https://deepcourse-epita.netlify.app/notebooks/VAE_Deepcourse.ipynb wget https://deepcourse-epita.netlify.app/notebooks/GAN_Deepcourse.ipynb wget https://deepcourse-epita.netlify.app/notebooks/Transformer_Deepcourse.ipynb wget https://deepcourse-epita.netlify.app/notebooks/MLP_Mixer_Deepcourse.ipynb {{< /highlight >}}

These notebooks are based on the material provided by Arthur Douillard.

Course Support

This course covers a wide range of mathematical tools for AI and science in general.We focus on Linear Algebra, Functional Analysis and Statistics.

• Rappels
• Gradients
• notes de cours du MVA

exercises

Here, you will find extra exercises for the exam:

• exercices supplémentaires (1) some exercises on confusion matrix and CLT.
• exercices supplémentaires (2) these exercises may be useful for the class of physics and in general to push the notions on gradients one step further.
• chain rule deux exercises sur la chain rule.
• exercices corriges supplementaires
• examen like et correction
• more examen-like

QCMs:

Practical Sessions

First, we need to install wget.

Install wget on Windows

google is your friend not me :)

Install wget on Linux

{{< highlight html >}} apt-get install wget {{< /highlight >}}

Install wget on MacOS

First, we need to install homebrew {{< highlight html >}} /bin/bash -c "$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/HEAD/install.sh)" {{< /highlight >}} Then use brew to install wget on your mac. {{< highlight html >}} brew install wget {{< /highlight >}}

Notebooks

Notebooks contain a small set of exercices to practice the different notions sutided in class. If you find any mistake please mail me.

{{< highlight html >}} wget https://edouardyvinec.netlify.app/centrale/notebooks/algèbre.ipynb wget https://edouardyvinec.netlify.app/centrale/notebooks/Classification.ipynb wget https://edouardyvinec.netlify.app/centrale/notebooks/EDO.ipynb wget https://edouardyvinec.netlify.app/centrale/notebooks/Fonctions_de_plusieurs_variables.ipynb wget https://edouardyvinec.netlify.app/centrale/notebooks/Probabilités.ipynb wget https://edouardyvinec.netlify.app/centrale/notebooks/TCL.ipynb wget https://edouardyvinec.netlify.app/centrale/notebooks/tests_d'hypothèses.ipynb {{< /highlight >}}

TP - Epidemic Modelisation

The base files for the TP are: {{< highlight html >}} wget https://edouardyvinec.netlify.app/centrale/tp/epidemic/exercice.py wget https://edouardyvinec.netlify.app/centrale/tp/epidemic/exercice_initial.py {{< /highlight >}} To visualize the expected results: {{< highlight html >}} wget https://edouardyvinec.netlify.app/centrale/tp/epidemic/SIR.png wget https://edouardyvinec.netlify.app/centrale/tp/epidemic/SIRVD.png wget https://edouardyvinec.netlify.app/centrale/tp/epidemic/SIRVDV.png {{< /highlight >}} Download the solution: {{< highlight html >}} wget https://edouardyvinec.netlify.app/centrale/tp/epidemic/main.py wget https://edouardyvinec.netlify.app/centrale/tp/epidemic/initial.py {{< /highlight >}}

TP - Solar System

The base files for the TP are: {{< highlight html >}} wget https://edouardyvinec.netlify.app/centrale/tp/solar_system/animate.py wget https://edouardyvinec.netlify.app/centrale/tp/solar_system/clear.py wget https://edouardyvinec.netlify.app/centrale/tp/solar_system/data.py wget https://edouardyvinec.netlify.app/centrale/tp/solar_system/main.py wget https://edouardyvinec.netlify.app/centrale/tp/solar_system/solarsystem_exercice.py wget https://edouardyvinec.netlify.app/centrale/tp/solar_system/utils.py {{< /highlight >}} To visualize the expected results: {{< highlight html >}} wget https://edouardyvinec.netlify.app/centrale/tp/solar_system/minisolar/minisolar.gif wget https://edouardyvinec.netlify.app/centrale/tp/solar_system/solar/solar.gif {{< /highlight >}} To convert a sequence of images to a gif: {{< highlight html >}} wget https://edouardyvinec.netlify.app/centrale/tp/gif/convertor.py {{< /highlight >}} Download the solution: to come {{< highlight html >}} wget https://edouardyvinec.netlify.app/centrale/tp/solar_system/solarsystem_sol.py {{< /highlight >}}

TP - Real Estate

The base files for the TP are: {{< highlight html >}} wget https://edouardyvinec.netlify.app/centrale/tp/real_estate/housing.csv wget https://edouardyvinec.netlify.app/centrale/tp/real_estate/main.py wget https://edouardyvinec.netlify.app/centrale/tp/real_estate/neural_network_exercice.py wget https://edouardyvinec.netlify.app/centrale/tp/real_estate/read_data.py {{< /highlight >}} Download the solution: to come

TP - Gaussian Mixture Model

The base files for the TP are: {{< highlight html >}} wget https://edouardyvinec.netlify.app/centrale/tp/mixture_of_gaussians/main.py wget https://edouardyvinec.netlify.app/centrale/tp/mixture_of_gaussians/ME_exercice_initial.py wget https://edouardyvinec.netlify.app/centrale/tp/mixture_of_gaussians/data.py {{< /highlight >}} To visualize the expected results: {{< highlight html >}} wget https://edouardyvinec.netlify.app/centrale/tp/mixture_of_gaussians/dataset.png {{< /highlight >}} Download the solution: to come

• slides
• notes

Practical Session

TP compression $\rightarrow$ [solution in torch and tensorflow]()

TP pose estimation

TP instance segmentation

• [x] A manager for json files, if you define the experiment parameters using json files, this manager may come handy. See the gitlab page or the pypi page.
• [x] A package for tensorflow models. This implements batch-normalization folding for keras models. You can check the gitlab page or the pypi page.
• [x] A basic toturial on deep neural networks compression. Check the gitlab page.
• [x] A simple package to clean the tensorflow Bert models (to have a cleaner load in the Netron). Check the gitlab page.