My road to learn python for deep learning.

My road to learn python for deep learning.

Preface:before I started to learn python for deep learning, the author is extremely familiar with the theory  of the deep learning. Meanwhile, the author is also familiar with two deep learning toolboxs (caffe and matconvnet).  Since I am very interested in python language, thus I decided to learn python for deep learning.

For deep learning theory, I recommend the following materials:

Michael Nielsen:Neural Network and Deep Learning, a online book(now updated to Chapter five).

Remarks: this book is awesome, I spent two days finishing this book. Interestingly, I got the feeling when I read the Pattern Recognition and Machine Learing(PRML).

Geoffrey E. Hinton:

Neural Network for Deep Learning, the course in coursera.

you can find many materials from the internet such as the deep learning course from stanford. I do not intend to mention them all.

What I have done:

To learn to use deep learning toolbox written in Python, such as Theano or Torch, you need to learn Python language well.If you have good knowledge to C++ and Matlab, you will find it not so hard to learn Python.

Python Skills, I recommend :

Google Python Class:,if you can access youtube, it will be good. you can watch the video.

Coursera course:

Numpy, Scipy, and matlabplot:

Python Imaging Library:

A tutorial post by RootOfTheNull,

A very good tutorial for Theano byAlec Radford:

Introduction to Deep Learning with Python

Until now, for python, I have read above-mentioned content. My future plan is to follow two projects,

1: Kaggle competions on Detecting the Local of Keypoints on Face Images.

The Link:

A very blog to Using CNN to detect facial keypoints tutorial:

2:Kaggle competions onPredict Ocean Health, one Plankton at a time

The Link:

The rank 1st method:

I have summitted one result to the facial keypoint detection and randed 3rd.

I will continue updating the content in future when I learn more. I do not check the language, pay attention to the contest and ignore the typos.