The goal of Keras is to make Neural Networks accessible and easy to code, running on top of state-of-the-art deep learning technologies like TensorFlow or CNTK. It focuses on friendliness, modularity and extensibility as principles that apply to either building a simple Neural Network or a complex Convolutional NN architecture.
In this talk we will present the Keras Sequential API, providing an intuitive methodology to create a model as a stack of layers. We will show a few interesting features like how it is possible to monitor your training process using TensorBoard or data pre-processing. Finally we will build a Neural Network and try to forecast stock prices.
Required audience experience
A general knowledge of basic ML principles (train vs test set etc). Some basic knowledge of Neural Networks. Some familiarity with an interactive programming language (Python, R).
Objective of the talk
Provide an introduction to Keras focused on its ease of use. The aim of a live coding exercise is to break the barrier between theory/static contents and practice.