benoitdescamps

def get_cost(target,Q,action_indices):
    """
    Cost-function of the Q-matrix attempting to approximate the reward-function
    :param tf.placeholder target: placeholder for the values of the registered rewards
    :param tf.placeholder Q: output of the Q-matrix the registered state
    :param tf.placeholder action_indices: placeholder for the indices of the registered actions
    :return: tf.tensor mean_squared error the reward vs Q-matrix
    """
    row_indices =  tf.range(tf.shape(action_indices)[0])
    full_indices = tf.stack([row_indices, action_indices], axis=1)

benoitdescamps

# Gradient Descent
optimizer = tf.train.GradientDescentOptimizer(learning_rate).minimize(cost)

benoitdescamps

class AddSign(optimizer.Optimizer):
    """Implementation of AddSign.
    See [Bello et. al., 2017](https://arxiv.org/abs/1709.07417)
    @@__init__
    """

    def __init__(self, learning_rate=1.001,alpha=0.01,beta=0.5, use_locking=False, name="AddSign"):
        super(AddSign, self).__init__(use_locking, name)
        self._lr = learning_rate

benoitdescamps

# This class defines the API to add Ops to train a model. 
from tensorflow.python.ops import control_flow_ops
from tensorflow.python.ops import math_ops
from tensorflow.python.ops import state_ops
from tensorflow.python.framework import ops
from tensorflow.python.training import optimizer
import tensorflow as tf

benoitdescamps

 def _create_slots(self, var_list):
     # Create slots for allocation and later management of additional 
     # variables associated with the variables to train.
     # for example: the first and second moments.
     '''
     for v in var_list:
     self._zeros_slot(v, "m", self._name)
     self._zeros_slot(v, "v", self._name)
     '''
 def _apply_dense(self, grad, var):

benoitdescamps

class TreeProperties(object):
    '''
    :param max_leafs: maximum number of leafs
    :param n_features: maximum number of feature available within the data
    :param n_classes: number of classes
    '''
    def __init__(self,max_depth,max_leafs,n_features,n_classes,regularisation_penality=10.,decay_penality=0.9):
        self.max_depth = max_depth
        self.max_leafs = max_leafs
        self.n_features = n_features

benoitdescamps

class Node(object):
    def __init__(self,id,depth,pathprob,tree):
        self.id = id
        self.depth = depth
        self.prune(tree)
        if self.isLeaf:
            self.W = tf.get_variable(...)
            self.b = tf.get_variable(...)
        else:
            self.W = tf.get_variable(...)

benoitdescamps

class SoftDecisionTree(object):
    def __init__(self, *args,**kwargs):
        self.params = TreeProperties(*args,**kwargs)

        self.loss = 0.0

        self.output = list()
        self.leafs_distribution = list()
    def build_tree(self):
        self.tf_X = tf.placeholder(tf.float32, [None, self.params.n_features])

benoitdescamps

cluster_spec = tf.train.ClusterSpec({'worker' : ['localhost:2222']})
server = tf.train.Server(cluster_spec)

benoitdescamps

def map_fun(args, ctx):
    try:
        import tensorflow as tf
        
        #utils
        from datetime import datetime
        import time
        import logging
        import numpy as np