Titanic-MTLDATA-Theano

Logistic Regression and a single layer neural network for the Kaggle Titanic Competition for MTLData - Oct.-22,2014

This IPython notebook is a continuation of [work (IPython-NB Link)] (http://nbviewer.ipython.org/github/aanchan/Titanic-MTLDATA/blob/master/TitanicPythonIntro.ipynb) from our previous meetup for the Kaggle Titanic Competition. The solution presented there was using Pandas+SciKit Learn. Click here for the Github page for that meetup.

Content

Data Preparation - A lot of this follows from work during our previous meetup, i.e : Data cleaning using Pandas.

Logistic Regression using Theano

Neural Network Training using Theano

Installing tools required for code in this tutorial run on your system.

1. Python
2. IPython (Optional since you could run the Python commands from the IPython 
   	   notebook on your native Python interpreter)
3. Numpy
4. Scipy
5. Pandas
6. Theano 

Installation methods for a scientific Python setup vary depending on the Operating System. Here is a great link on completing a setup in Python for scientific purposes.

Installation instructions for Theano are available from the Theano website

The Kaggle Titanic Challenge

Read about it here

Introduction to Python

Course from Coursera. This does not require one to download and install Python. They have a version for the course that runs off the browser interactively.

The best intro I think, from Python Docs

Introduction to the Numpy module in Python

The Tentative Numpy Tutorial is a good place to start.

Introduction to Pandas

The Python Pandas Cookbook Lecture Series on Youtube by Alfred Essa is a good place to start. Specifically to load our Titanic data set Alfred Essa talks about it here in Lesson 1.2.

Introduction to Theano

While the great tutorial webpages appear on the Theano website, a companion IPython notebook with similar content, and especially a great intro code to Neural Networks is available here.

Introduction to Logistic Regression

A Simple Explanation from Duke Medicine

Logistic Regression for Classification

Introduction to Neural Networks

Logistic and Softmax Regression by Prof. Andrew Ng at Stanford.

A clear (and correct) introduction to Neural Networks by Prof. Andrew Ng at Stanford.