Skip to content

abhishekraok/LayeredNeuralNetwork

Repository files navigation

LayeredNeuralNetwork

A transfer learning Neural Network that uses features learnt from past training.

Build Status

This is implementation of Layered Neural Network (LNN) first proposed in this MS Thesis

The difference between normal classifier and this one is this is not reset after every training. Here every training task has a label (like class name). Once a mapping from input to output is learnt, this is used as a feature for the subsequent task. LNN structure

API

  • Fit(X,Y, label) : Trains the model using data X for class named "label". Y is binary indicating presence/absence of class. X is a numpy matrix of size (samples, features), Y is a numpy array of integers with values 0,1. label is a string.
  • Predict(X, label) : Predicts presence or absence of class "label" in X.
  • Identify(X) : Guesses the best class for X

To Do

Done

  • naming nodes incrementally

About

Neural Network that uses features learnt from past training.

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published