Tenssorflow tutorial for MLSS Algiers
This tutorial teaches the main concepts of tensorflow, necessary to extend and develop new machine learning models and algorithms. The tutorial is split into 4 parts.
- Part 1 introduces the basics or mechanics of tensorflow
- In part 2, you will implement our first Machine Learning model and training algorithm with the low-level tf API.
- In part 3, we introduce very practical and useful high-level APIs to facilitate implementation and debugging.
- In part 4, you will adapt your model from classification to a regression problem.
We only need tensorflow, numpy, matplotlib, and jupyter notebook, preferably with python3. In this tutorial, we work with simple models and toy data, so we don't need GPU support.
- Install python3 and pip
- Recommended: virtualenv with virtualenvwrapper (to create isolated environment with python packages for this tutorial).
sudo pip install virtualenv virtualenvwrapper
add the following lines to your ~/.bashrc or ~/.zshrc or ~/.bash_profile (depends what you are using)
export WORKON_HOME=$HOME/.virtualenvs export VIRTUALENVWRAPPER_PYTHON=/usr/local/bin/python3 export VIRTUALENVWRAPPER_VIRTUALENV=/usr/local/bin/virtualenv export VIRTUALENVWRAPPER_VIRTUALENV_ARGS='--no-site-packages'
mkvirtualenv tf_tutorial --python=python3
- pip install matplotlib numpy tensorflow jupyter notebook
Alternative - Execute on google colab:
You can run your code on some google machines for free.
Go to https://colab.research.google.com and sign in with your google account (you need one to use colab)
File --> open notebook --> https://github.com/richardk53/tf_tutorial_algiers.git