Build a sampler which only steps once.

Build a sampler which only steps once.

Build a sampler which only steps once. Typically it only works for one word a time?

Build a sampler which only steps once. Typically it only works for one word a time?

Expects a binary class matrix instead of a vector of scalar classes

Checks that the subtree rooted at t is a valid BST and all keys are between (lokey, hikey).

build the hidden action prediction.

build the hidden action prediction.

build the training function here <:::>

Delete this node from the tree.

display a dialogue session

display a dialogue session

build the evaluation function for valid/testing Note that we need multiple sampling for this!

Return the node for key t if it is in this tree, or None otherwise.

Computes the KL-divergence of Gaussian variables from the standard one. Given two variable ``mean`` representing :math:`\\mu` and ``ln_var``

:param action: action vector to guide the question. If None, use a Gaussian to simulate the action. :return: q

:param action: action vector to guide the question. If None, use a Gaussian to simulate the action. :return: q

:param action: action vector to guide the question. If None, use a Gaussian to simulate the action. :return: q

For some models (such as RNNs) you want a way of being able to mark some output data-points as "masked", so they are not used in futu

classification=True overrides output_shape (i.e. output_shape is set to (1,)) and the output consists in integers in [0, nb_c

Reference: Glorot & Bengio, AISTATS 2010

Reference: He et al., http://arxiv.org/abs/1502.01852

Insert key t into the subtree rooted at this node (updating subtree size).

Reference: LeCun 98, Efficient Backprop http://yann.lecun.com/exdb/publis/pdf/lecun-98b.pdf

The function returns the variable that is passed in, so all types work