This is a tutorial on Tensorflow that starts from its basics (constants and variables) to implementing different kinds of neural network architectures. Keras is a library built on tensorflow and theano and allows users to create neural networks in an abstract form (i.e. create a neural network by adding different kinds of layers without defining their operations). The tutorial contains same examples implemented in Keras so that it easy to gauge which libarary is more preferable according to the requirements. Most python libraries are based on NumPy & SciPy. The tutorial uses the python Pandas framework with these libraries as a way to manage data.
The tutorial is inspired by Tensorflow-examples in structure and Abhay Kashyap tutorial on Tensorflow.
An introduction to Tensorflow and a look at MNIST, a standard dataset that is used in Deep Learning. Also, added are list of links to good materials available online.
- Resource Links - A list of some of he popular online resources to study deep learning / neural networks.
- Tensorflow Basics - An overview of tensorflow basics and operations.
An introduction to basic machine learnin models and how they can be implemented in various libraries. Each notebook provides the setup for the problem, the data and how the model is trained.
An introduction to Feed Forward Networks (or Multilayer Perceptrons).
Different convolutional neural network architectures, with their explanations.