A factory for weight initializers. Parameters ---------- act_fn_str : str The string representation for

A WaveNet-like residual block with causal dilated convolutions. .. code-block:: text *Skip path in* >------------------

A ResNet-like "identity" shortcut module. Notes ----- The identity module enforces `same` padding during each convol

A ResNet-like "convolution" shortcut module. Notes ----- In contrast to :class:`SkipConnectionIdentityModule`, the a

A single bidirectional long short-term memory (LSTM) layer. Parameters ---------- n_out : int The dimens

A mutli-headed attention module. Notes ----- Multi-head attention allows a model to jointly attend to information fr

An abstract base class for all Wrapper instances

Initialize the ActivationBase object

A logistic sigmoid activation function.

r""" A Gaussian error linear unit (GELU). [*]_ Notes ----- A ReLU alternative. GELU weights inputs by their value, ra

A hyperbolic tangent activation function.

An affine activation function. Parameters ---------- slope: float Activation slope. Default is 1.

Identity activation function. Notes ----- :class:`Identity` is syntactic sugar for :class:`Affine` with slop

An exponential (base e) activation function

A "hard" sigmoid activation function. Notes ----- The hard sigmoid is a piecewise linear approximation of the logist

A softplus activation function. Notes ----- In contrast to :class:`ReLU`, the softplus activation is differentiable

An abstract base class inherited by all neural network layers

r""" A single "attention head" layer using a dot-product for the scoring function. Notes ----- The equations for a do

An "addition" layer that returns the sum of its inputs, passed through an optional nonlinearity. Parameters --------

A multiplication layer that returns the *elementwise* product of its inputs, passed through an optional nonlinearity. Parame

Flatten a multidimensional input into a 2D matrix. Parameters ---------- keep_dim : {'first', 'last', -1}

A batch normalization layer for two-dimensional inputs with an additional channel dimension. Notes ----- Bat

A batch normalization layer for 1D inputs. Notes ----- BatchNorm is an attempt address the problem of internal covar

A layer normalization layer for 2D inputs with an additional channel dimension. Notes ----- In contrast to :

A layer normalization layer for 1D inputs. Notes ----- In contrast to :class:`BatchNorm1D`, the LayerNorm layer calc

An embedding layer. Notes ----- Equations:: Y = W[x] NB. This layer must be the first in a neu

r""" A fully-connected (dense) layer. Notes ----- A fully connected layer computes the function .. math::

r""" A softmax nonlinearity layer. Notes ----- This is implemented as a layer rather than an activation primarily

r""" A sparse Erdos-Renyi layer with evolutionary rewiring via the sparse evolutionary training (SET) algorithm. Notes

Apply a one-dimensional convolution kernel over an input volume. Notes ----- Equations:: out = act_fn(p

Apply a two-dimensional convolution kernel over an input volume. Notes ----- Equations:: out = act_fn(p

A single two-dimensional pooling layer. Parameters ---------- kernel_shape : 2-tuple The dimension of a

Apply a two-dimensional "deconvolution" to an input volume. Notes ----- The term "deconvolution" in this context doe

r""" A single step of a vanilla (Elman) RNN. Notes ----- At timestep `t`, the vanilla RNN cell computes .. m

A single step of a long short-term memory (LSTM) RNN. Notes ----- Notation: - ``Z[t]`` is the input to eac

A single vanilla (Elman)-RNN layer. Parameters ---------- n_out : int The dimension of a single hidden s

A single long short-term memory (LSTM) RNN layer. Parameters ---------- n_out : int The dimension of a s

Returns a fixed learning rate, regardless of the current step. Parameters ---------- initial_lr : float

An exponential learning rate scheduler. Notes ----- The exponential scheduler decays the learning rate by `decay` ev

The Noam learning rate scheduler, originally used in conjunction with the Adam optimizer in [1]. Notes -----

The Davis King / DLib learning rate scheduler. Notes ----- The KingScheduler computes the probability that the slope

An abstract base class for all Optimizer objects. This should never be used directly.

An AdaGrad optimizer. Notes ----- Weights that receive large gradients will have their effective learning ra

RMSProp optimizer. Notes ----- RMSProp was proposed as a refinement of :class:`AdaGrad` to reduce its aggres

Adam (adaptive moment estimation) optimization algorithm. Notes ----- Designed to combine the advantages of :class:`

Wasserstein generative adversarial network with gradient penalty. Parameters ---------- g_hidden : int T

A variational autoencoder (VAE) with 2D convolutional encoder and Bernoulli input and output units. Notes -----

A word2vec model supporting both continuous bag of words (CBOW) and skip-gram architectures, with training via noise contrastive

A squared-error / `L2` loss. Notes ----- For real-valued target **y** and predictions :math:`\hat{\mathbf{y}}`, the

A cross-entropy loss. Notes ----- For a one-hot target **y** and predicted class probabilities :math:`\hat{\

The loss function for a Wasserstein GAN [*]_ [*]_ with gradient penalty. Notes ----- Assuming an optimal critic, min

A noise contrastive estimation (NCE) loss function. Notes ----- Noise contrastive estimation is a candidate sampling

r""" A cross-entropy method agent. Notes ----- The cross-entropy method [1]_ [2]_ agent only operates on ``envs`` wit

A Monte-Carlo learning agent trained using either first-visit Monte Carlo updates (on-policy) or incremental weighted importance samp

r""" A temporal difference learning agent with expected SARSA (on-policy) [3]_ or TD(0) `Q`-learning (off-policy) [4]_ updates.

r""" A Dyna-`Q` / Dyna-`Q+` agent [5]_ with full TD(0) `Q`-learning updates via prioritized-sweeping [6]_ . Notes ---

An object to facilitate agent training and evaluation. Parameters ---------- agent : :class:`AgentBase` instance

A multi-armed bandit where each arm is associated with a different multinomial payoff distribution. Parameters -----

A multi-armed bandit where each arm is associated with an independent Bernoulli payoff distribution. Parameters ----

A multi-armed bandit that is similar to :class:`BernoulliBandit`, but instead of each arm having a fixed payout of 1, the pay

A weighted graph shortest path problem formulated as a multi-armed bandit. Notes ----- Each arm corresponds

A contextual version of :class:`BernoulliBandit` where each binary context feature is associated with an independent Bernoulli payoff

r""" A contextual linear multi-armed bandit. Notes ----- In a contextual linear bandit the expected payoff of an arm

A simple base class for multi-armed bandit policies

r""" An epsilon-greedy policy for multi-armed bandit problems. Notes ----- Epsilon-greedy policies greedily select th

r""" A conjugate Thompson sampling [1]_ [2]_ policy for multi-armed bandits with Bernoulli likelihoods. Notes -----

A disjoint linear UCB policy [*]_ for contextual linear bandits. Notes ----- LinUCB is only defined for :class:`Cont

An object to facilitate multi-armed bandit training, comparison, and evaluation.

A generic node object for holding entries in :class:`PriorityQueue`

A priority queue implementation using a binary heap. Notes ----- A priority queue is a data structure useful for sto

A ball tree data structure. Notes ----- A ball tree is a binary tree in which every node defines a `D`-dimen

Sample from an arbitrary multinomial PMF over the first `N` nonnegative integers using Vose's algorithm for the alias method.

A dictionary subclass which returns the key value if it is not in the dict. Parameters ---------- encoder :

The linear (i.e., dot-product) kernel. Notes ----- For input vectors :math:`\mathbf{x}` and :math:`\mathbf{y}`, the

The degree-`d` polynomial kernel. Notes ----- For input vectors :math:`\mathbf{x}` and :math:`\mathbf{y}`, the polyn

A class for initializing learning rate schedulers. Valid inputs are: (a) __str__ representations of `KernelBase` instances

A generic directed graph object. Parameters ---------- V : list A list of vertex IDs. E : list o

A generic undirected graph object. Parameters ---------- V : list A list of vertex IDs. E : list

A Gaussian Process (GP) regression model. .. math:: y \mid X, f &\sim \mathcal{N}( [f(x_1), \ldots, f(x_n)], \\alpha

A Nadaraya-Watson kernel regression model. Notes ----- The Nadaraya-Watson regression model is .. math::

A `k`-nearest neighbors (kNN) model relying on a ball tree for efficient computation. Parameters ----------

r""" A simple hidden Markov model with multinomial emission distribution. Parameters ---------- A : :py:class:`ndarra

Convert between category labels and their one-hot vector representations. Parameters ---------- categories :

Feature-wise standardization for vector inputs. Notes ----- Due to the sensitivity of empirical mean and standard de

Convert a collection of features to a fixed-dimensional matrix using the hashing trick. Notes ----- Uses the

A byte-pair encoder for sub-word embeddings. Notes ----- Byte-pair encoding [1][2] is a compression algorithm that i

r""" An object for compiling and encoding the term-frequency inverse-document-frequency (TF-IDF) representation of the tokens in a

An object for compiling and encoding the unique tokens in a text corpus. Parameters ---------- lowercase : bool

r""" Approximately factor a real-valued matrix using regularized alternating least-squares (ALS). Notes -----