This loss is similar to the native cosine_proximity loss from Keras but it differs by the fact that only the two first components of the two vect

Get the discount factor

Get the epsilon

Get all parameters used by the learning algorithm Returns ------- Values of the parameters: list of numpy arrays

Get all parameters used by the learning algorithm Returns ------- Values of the parameters: list of numpy arrays

This returns whether the environment reached a terminal state after the last transition (i.e. whether the last transition that occurred was

Tell whether the environment reached a terminal state after the last transition (i.e. the last transition that occured was terminal).

Tell whether the environment reached a terminal state after the last transition (i.e. the last transition that occured was terminal).

Get the learning rate

Modified mean square error that clips

Provides the bounds on the action space Returns ------- bounds on the action space

Possibility to override the chosen action. This possibility should be used on the signal OnActionChosen.

Get the q values for one pseudo-state (without planning) Arguments --------- state_val : array of objects (or list of object

Get the q value for one pseudo-state

Returns: current observation (list of k elements)

Resets the environment for a new episode. Parameters ----------- mode : int -1 is for the training phase, others

Set all parameters used by the learning algorithm Arguments --------- list_of_values : list of numpy arrays lis

Set all parameters used by the learning algorithm Arguments --------- list_of_values : list of numpy arrays

Activate controller

Set the discount factor

Set the epsilon

Set the learning rate for the gradient descent

Setting the learning rate NB: The learning rate has usually to be set in the optimizer, hence this function should be overridden. Oth

Set values into the network Parameters ----------- fname : string Name of the file where the values are

Plot of the low-dimensional representation of the environment built by the model

Plot of the low-dimensional representation of the environment built by the model

This function is called at every PERIOD_BTW_SUMMARY_PERFS. Parameters ----------- test_data_set

Get all terminals currently in the replay memory, ordered by time where they were observed. terminals[i] is True if actions()[i] lead

This function selects a random batch of data (with self._dataset.randomBatch) and performs a training iteration (with self._learning

Train the Q-network from one batch of data. Parameters ----------- states_val : numpy array of objects E

Train the actor-critic algorithm from one batch of data. Parameters ----------- states_val : numpy array of obj

This method performs the training step (e.g. using Bellman iteration in a deep Q-network) for one batch of tuples.