Run a whole iteration and updates the parameters of meta-learner. :param inner_objective_feed_dicts: feed dictionary for the Lower-L

reuses defined convolutional networks with new input and update the output results :param new_input: new input with same shape as the

Generates a `Datasets` object from a list. :param list_of_datasets: list containing from one to three dataset :return:

Remove tensors in lst from collection given by key

Generator of datasets :param rand: random seed, state or None :param count: number of datasets to generate :param ba

The `dynamics` contains a list of var_and_dynamics where var are both variables in `var_list` and also additional state (auxiliary) variables

Remove tensors in lst from collection given by key

Make sure that `obj` is a tuple or a list and eventually converts it into a list with a single element :param obj: :return: A `tuple` or

:param boml_opt: instance of modified optimizers in the `optimizer` module :param loss_inner: Lower-Level objectives :param l

Generates and returns a single Datasets (possibly composed by training, validation and test sets) according to args and kwargs

Utility methods for getting a `RandomState` object. :param rand: rand can be None (new State will be generated), np.rand

Function that redivides datasets. Can be use also to shuffle or filter or map examples. :param rand: :param balance_classes: # TODO RICC

Assuming that the datasets have same structure, stacks data, targets and other info :param datasets: :return: stacked datase

To be called when no more dynamics or problems will be added, computes the updates for the outer parameters. This behave nicely with

Apply gradients to variables. This is the second part of `minimize()`. It returns an `Operation` that applies gradients. Arg

This method has to be called for every experiment and takes responsibility for defining task-specific modules and inner optimizers.

Perform, conv, batch norm, nonlinearity, and max pool :param boml_net: instance of BOMLNet :param cweight: parameter of convolutional filter

:param primary_outerparameter: the primary outerparameters used to create the task-specific parameters :return: dictionary to

After construction of neural networks, solutions to lower level problems should be regulated in LL_Problem. :param inner_objective: l

This method must be called once at first to build meta modules and initialize meta parameters and neural networks. :param _input: org

Set the outer optimization problem and the descent procedure for the optimization of the outer parameters. Can be called at least onc

:return: compute the shape of input according to the data formats

:return: compute the shape of output according to the data formats

:return: generator of read value op for the state variables

sum elements in two lists by sequence :return: new list that sum elements in two lists by sequence

Makes sure `obj` is a list or otherwise converts it to a list with a single element. :param obj: :return: A `list`

Generates and stores example visiting scheme, as a numpy array of integers. :return: self

:param d_set: instance of dataset :return: inputs of datasets

:param what: input :return: return the scalar of input if the length of input equals 1.

:param _example: :param _size: :return: performs concatenation

for save setting purposes, does not save the actual data :return:

Builds a feed dictionary of (past) states

:param seq: sequence :param dimension: the dimension of output :return: output of one-hot encoding

Vstack that considers sparse matrices :param lst: :return:

:param _input: original input :param dim_output: dimension of output :param name: scope of meta-learner :param outer_

:param _input: original input :param dim_output: dimension of output :param name: scope of meta-learner :param outer_

:param data: Numpy array containing data :param target: Numpy array containing targets :param sample_info: either an array o

Creates one hyper-gradient as a variable. doo_dhypers: initialization, that is the derivative of the outer objective w.r.t this hype

Creates one hyper-gradient as a variable. doo_dhypers: initialization, that is the derivative of the outer objective w.r.t this hype

_forward() uses defined convolutional neural networks with initial input :return:

Replace chars that are not accepted by tensorflow namings (eg. variable_scope) :param _str: :return:

May override default variable scope context. :param reuse: :return:

Make sure that `obj` is a tuple or a list and eventually converts it into a list with a single element :param obj: :return: A `tuple` or

:param inner_optimizer: the instance of modified optimizer :param loss_inner: inner objective :param loss_outer: outer object

:param loss_inner: LL objectives :param param_dict: dictionary of general parameters for different algorithms :param var_lis

:param xx: input for the i-th block :param i: i-th block :param j: the j-th convolution block :return: the ouput of

:param X: sparse matrix :return: sparse tensor of X

:param name: optional name for this supplier :param x: placeholder for independent variable :param y: placeholder for depend

:param var_collections: name of collections to store the created variables. :return: dictionary to index the created variables.

:param var_collections: name of collections to store the created variables. :return: dictionary to index the created variables.

:param var_name: name of the variable :param scope: name scope :return: variable with the name in the scope

:param lst: original list :return: flattened list

Generates a batch of Datasets

:param n_examples: number of examples :param labels: paired labels :param forbidden_indices: :return:

:param d_set: instance of dataset :return: labels of datasets

Judge whether the num is integer :param num: :return: BOOLEAN