Introduction to Neural Networks and TensorFlow with guided exercises.

The project is based on a Python 2 Anaconda distribution. Download it from the main website if needed.

It is recommended to create a separate environment to work with TensorFlow to avoid any version problem in your main Python environment. Source the environment once created.

conda create --name=tensorflow_env python=2.7
source activate tensorflow_env

Before downloading TensorFlow, make sure you don't have a installation of the protobuf library, which has some version problems with TensorFlow. Once done, you can download TensorFlow.

pip uninstall protobuf
export TF_BINARY_URL=https://storage.googleapis.com/tensorflow/mac/tensorflow-0.9.0rc0-py2-none-any.whl
pip install --upgrade $TF_BINARY_URL

git clone https://github.com/ybenoit/hands-on-tensorflow.git
cd hands-on-tensorflow

cd hands-on-tensorflow
py.test hands_on_tensorflow/solutions/mnist_1_softmax_regression/tests/

To be able to work with TensorFlow in an IDE like PyCharm, you must specify the Python Interpreter corresponding to your TensorFlow conda environment.

• Open PyCharm
• PyCharm -> Preferences -> Project Interpreter
• Select the interpreter corresponding to your TensorFlow anaconda environment (Ex: ~/anaconda/envs/tensorflow/bin/python)

The project is organized as follow:

• The tutorials/exercices module is your working module
• The tutorials/solutions module contains the solutions of the exercices

To complete the exercices, follow the instructions in the presentation (prez directory).