Learn Deep Learning in One Day - Keras Demo

This is my implementation of demo code in 一天搞懂深度學習.

Description

In this demo, the nerual network is trained using mnist data to recognize numbers.

The initial model uses MSE as cost function and sigmoid as activation function which would get a bad performance when only 10,000 data was used.

The following methods (taught in the 一天搞懂深度學習 lesson) are used to enhance accuracy and avoid overfitting.

You can simply open the Demo.ipynb on GitHub or through jupyte nbviewer(recommended).

This ipython notebook is also organized as slides that can be used with default jupyter nbcoverter or RISE.