hula.DeepR.RecurrentLayer

hula/DeepR.py

import hula.DeepR.RecurrentLayer

from hula.DeepR import RecurrentLayer

RecurrentLayer( inputs, outputs, activation_func )

Generated recurrent nodes with the activation function activation_func with the size of inputs byoutputs

RecurrentLayer.activate( X )

Feeds X through it's nodes, and returns the output, while updating the recurrent state to the output

RecurrentLayer.randomAct( alpha )

Generates a random action to apply to its weights, as well as it's recurrent weights, then stores it for scoring later

RecurrentLayer.score( score )

Applies score to the stored actions

RecurrentLayer.simplify( threshold, min_score )

Groups together states with threshold distance from each other, then only allows states that have a score higher than min_score to rejoin the State-Action Tree

RecurrentLayer.train( alpha )

Applies the best selected actions according to their state's distance from the current state as well as their score, then only applies alpha percentage of the action.