Add Initializers

[jramapuram]

Currently Nodes are defined to have spreads:

            LinearNode(const int inputSize, const int outputSize,
                       float spread = 0.05,
                       const char *name="none") :

This needs to be refactored to provide an initializations class. Examples of initializations are :

1. Le-Cun Normal
2. Glorot Uniform

[pavanky]

Erm.. Googling these terms seems to produce a lot of noise. Can you link to proper documentation ?

[jramapuram]

Yea, sorry. Keras has some links to the papers as well as the implementations: https://github.com/fchollet/keras/blob/master/keras/initializations.py

[pavanky]

This can be implemented by adding a few functions in Types.hpp.

We'll need to implement the following:

• [x] The current nn.weight should be renamed to nn.randomUniform.
• [x] nn.randomNormal for normal
• [x] nn.constant similar to af::constant
• [x] nn.identity
• [x] nn.lecunNormal
• [x] nn.lecunUniform
• [x] nn.glorotNormal
• [x] nn.glorotUniform

The links for lecun and glorot can be found here: https://github.com/fchollet/keras/blob/master/keras/initializers.py