rodrigogiraoserrao / ANNAPL

Artificial Neural Networks framework written in APL.

ANNAPL

Artificial Neural Networks in APL.

How To Use

Clone this repository to a directory, say c:\tmp, then fire up your interpreter and create a link with the folder as a namespace

      ]create ANNAPL c:\tmp\ANNAPL
Linked: #.ANNAPL ←→ c:\tmp\ANNAPL

After the link is successfully created, go ahead and run one of the examples:

      ANNAPL.examples.corner ⍬

Examples

Take a look at the examples to see how to use the networks. In particular, the corner.aplf is a simple tradfn with a couple of lines exhibiting how to do the simplest things.

Components Available

To build a neural network you will need:

• an array of layers;
  • each of which needs an activation function; and
  • a shape.
• a loss function; and
• a learning rate.

Use Layer to build layers and Net to build a network.

Use _FeedForward to feed inputs to the network and use _Train to train the network on inputs and expected targets.

Activation Functions

Here are the activation functions implemented in ActivationFns:

• Id – identity function
• ReLU – rectified linear unit
• LeakyReLU – leaky rectified linear unit (parametrised)
• ELU – exponential linear unit
• Sigmoid
• Tanh – Hyperbolic tangent
• ArcTan – Arc Tangent

Loss Functions

Here are the loss functions implemented in LossFns:

• BCELoss – binary cross entropy loss
• CrossEntropyLoss
• L1Loss
• MSELoss – mean squared error loss