\section{R\+NN$<$ Output\+Layer\+Type, Initialization\+Rule\+Type, Custom\+Layers $>$ Class Template Reference} \label{classmlpack_1_1ann_1_1RNN}\index{R\+N\+N$<$ Output\+Layer\+Type, Initialization\+Rule\+Type, Custom\+Layers $>$@{R\+N\+N$<$ Output\+Layer\+Type, Initialization\+Rule\+Type, Custom\+Layers $>$}} Implementation of a standard recurrent neural network container. Inheritance diagram for R\+NN$<$ Output\+Layer\+Type, Initialization\+Rule\+Type, Custom\+Layers $>$\+: \nopagebreak \begin{figure}[H] \begin{center} \leavevmode \includegraphics[width=207pt]{classmlpack_1_1ann_1_1RNN__inherit__graph} \end{center} \end{figure} \subsection*{Public Types} \begin{DoxyCompactItemize} \item using \textbf{ Network\+Type} = \textbf{ R\+NN}$<$ Output\+Layer\+Type, Initialization\+Rule\+Type, Custom\+Layers... $>$ \begin{DoxyCompactList}\small\item\em Convenience typedef for the internal model construction. \end{DoxyCompactList}\end{DoxyCompactItemize} \subsection*{Public Member Functions} \begin{DoxyCompactItemize} \item \textbf{ R\+NN} (const size\+\_\+t rho, const bool single=false, Output\+Layer\+Type output\+Layer=Output\+Layer\+Type(), Initialization\+Rule\+Type initialize\+Rule=Initialization\+Rule\+Type()) \begin{DoxyCompactList}\small\item\em Create the \doxyref{R\+NN}{p.}{classmlpack_1_1ann_1_1RNN} object. \end{DoxyCompactList}\item \textbf{ $\sim$\+R\+NN} () \begin{DoxyCompactList}\small\item\em Destructor to release allocated memory. \end{DoxyCompactList}\item {\footnotesize template$<$class Layer\+Type , class... Args$>$ }\\void \textbf{ Add} (Args... args) \item void \textbf{ Add} (\textbf{ Layer\+Types}$<$ Custom\+Layers... $>$ layer) \item double \textbf{ Evaluate} (const arma\+::mat \¶meters, const size\+\_\+t begin, const size\+\_\+t batch\+Size, const bool deterministic) \begin{DoxyCompactList}\small\item\em Evaluate the recurrent neural network with the given parameters. \end{DoxyCompactList}\item double \textbf{ Evaluate} (const arma\+::mat \¶meters, const size\+\_\+t begin, const size\+\_\+t batch\+Size) \begin{DoxyCompactList}\small\item\em Evaluate the recurrent neural network with the given parameters. \end{DoxyCompactList}\item {\footnotesize template$<$typename Grad\+Type $>$ }\\double \textbf{ Evaluate\+With\+Gradient} (const arma\+::mat \¶meters, const size\+\_\+t begin, Grad\+Type \&gradient, const size\+\_\+t batch\+Size) \begin{DoxyCompactList}\small\item\em Evaluate the recurrent neural network with the given parameters. \end{DoxyCompactList}\item void \textbf{ Gradient} (const arma\+::mat \¶meters, const size\+\_\+t begin, arma\+::mat \&gradient, const size\+\_\+t batch\+Size) \begin{DoxyCompactList}\small\item\em Evaluate the gradient of the recurrent neural network with the given parameters, and with respect to only one point in the dataset. \end{DoxyCompactList}\item size\+\_\+t \textbf{ Num\+Functions} () const \begin{DoxyCompactList}\small\item\em Return the number of separable functions (the number of predictor points). \end{DoxyCompactList}\item const arma\+::mat \& \textbf{ Parameters} () const \begin{DoxyCompactList}\small\item\em Return the initial point for the optimization. \end{DoxyCompactList}\item arma\+::mat \& \textbf{ Parameters} () \begin{DoxyCompactList}\small\item\em Modify the initial point for the optimization. \end{DoxyCompactList}\item void \textbf{ Predict} (arma\+::cube predictors, arma\+::cube \&results, const size\+\_\+t batch\+Size=256) \begin{DoxyCompactList}\small\item\em Predict the responses to a given set of predictors. \end{DoxyCompactList}\item const arma\+::cube \& \textbf{ Predictors} () const \begin{DoxyCompactList}\small\item\em Get the matrix of data points (predictors). \end{DoxyCompactList}\item arma\+::cube \& \textbf{ Predictors} () \begin{DoxyCompactList}\small\item\em Modify the matrix of data points (predictors). \end{DoxyCompactList}\item void \textbf{ Reset} () \begin{DoxyCompactList}\small\item\em Reset the state of the network. \end{DoxyCompactList}\item void \textbf{ Reset\+Parameters} () \begin{DoxyCompactList}\small\item\em Reset the module information (weights/parameters). \end{DoxyCompactList}\item const arma\+::cube \& \textbf{ Responses} () const \begin{DoxyCompactList}\small\item\em Get the matrix of responses to the input data points. \end{DoxyCompactList}\item arma\+::cube \& \textbf{ Responses} () \begin{DoxyCompactList}\small\item\em Modify the matrix of responses to the input data points. \end{DoxyCompactList}\item const size\+\_\+t \& \textbf{ Rho} () const \begin{DoxyCompactList}\small\item\em Return the maximum length of backpropagation through time. \end{DoxyCompactList}\item size\+\_\+t \& \textbf{ Rho} () \begin{DoxyCompactList}\small\item\em Modify the maximum length of backpropagation through time. \end{DoxyCompactList}\item {\footnotesize template$<$typename Archive $>$ }\\void \textbf{ serialize} (Archive \&ar, const unsigned int) \begin{DoxyCompactList}\small\item\em Serialize the model. \end{DoxyCompactList}\item void \textbf{ Shuffle} () \begin{DoxyCompactList}\small\item\em Shuffle the order of function visitation. \end{DoxyCompactList}\item {\footnotesize template$<$typename Optimizer\+Type , typename... Callback\+Types$>$ }\\double \textbf{ Train} (arma\+::cube predictors, arma\+::cube responses, Optimizer\+Type \&optimizer, Callback\+Types \&\&... callbacks) \begin{DoxyCompactList}\small\item\em Train the recurrent neural network on the given input data using the given optimizer. \end{DoxyCompactList}\item {\footnotesize template$<$typename Optimizer\+Type = ens\+::\+Standard\+S\+GD, typename... Callback\+Types$>$ }\\double \textbf{ Train} (arma\+::cube predictors, arma\+::cube responses, Callback\+Types \&\&... callbacks) \begin{DoxyCompactList}\small\item\em Train the recurrent neural network on the given input data. \end{DoxyCompactList}\item {\footnotesize template$<$typename Optimizer\+Type $>$ }\\std\+::enable\+\_\+if$<$ Has\+Max\+Iterations$<$ Optimizer\+Type, size\+\_\+t \&(Optimizer\+Type\+::$\ast$)()$>$\+::value, void $>$\+::type \textbf{ Warn\+Message\+Max\+Iterations} (Optimizer\+Type \&optimizer, size\+\_\+t samples) const \begin{DoxyCompactList}\small\item\em Check if the optimizer has Max\+Iterations() parameter, if it does then check if it\textquotesingle{}s value is less than the number of datapoints in the dataset. \end{DoxyCompactList}\item {\footnotesize template$<$typename Optimizer\+Type $>$ }\\std\+::enable\+\_\+if$<$ !Has\+Max\+Iterations$<$ Optimizer\+Type, size\+\_\+t \&(Optimizer\+Type\+::$\ast$)()$>$\+::value, void $>$\+::type \textbf{ Warn\+Message\+Max\+Iterations} (Optimizer\+Type \&optimizer, size\+\_\+t samples) const \begin{DoxyCompactList}\small\item\em Check if the optimizer has Max\+Iterations() parameter, if it doesn\textquotesingle{}t then simply return from the function. \end{DoxyCompactList}\end{DoxyCompactItemize} \subsection{Detailed Description} \subsubsection*{template$<$typename Output\+Layer\+Type = Negative\+Log\+Likelihood$<$$>$, typename Initialization\+Rule\+Type = Random\+Initialization, typename... Custom\+Layers$>$\newline class mlpack\+::ann\+::\+R\+N\+N$<$ Output\+Layer\+Type, Initialization\+Rule\+Type, Custom\+Layers $>$} Implementation of a standard recurrent neural network container. \begin{DoxyTemplParams}{Template Parameters} {\em Output\+Layer\+Type} & The output layer type used to evaluate the network. \\ \hline {\em Initialization\+Rule\+Type} & Rule used to initialize the weight matrix. \\ \hline \end{DoxyTemplParams} Definition at line 45 of file rnn.\+hpp. \subsection{Member Typedef Documentation} \mbox{\label{classmlpack_1_1ann_1_1RNN_a8b97fdc6b24723019bcab248fe7db76a}} \index{mlpack\+::ann\+::\+R\+NN@{mlpack\+::ann\+::\+R\+NN}!Network\+Type@{Network\+Type}} \index{Network\+Type@{Network\+Type}!mlpack\+::ann\+::\+R\+NN@{mlpack\+::ann\+::\+R\+NN}} \subsubsection{Network\+Type} {\footnotesize\ttfamily using \textbf{ Network\+Type} = \textbf{ R\+NN}$<$Output\+Layer\+Type, Initialization\+Rule\+Type, Custom\+Layers...$>$} Convenience typedef for the internal model construction. Definition at line 51 of file rnn.\+hpp. \subsection{Constructor \& Destructor Documentation} \mbox{\label{classmlpack_1_1ann_1_1RNN_aa07a59fdcbe988264200c8f593c73bbf}} \index{mlpack\+::ann\+::\+R\+NN@{mlpack\+::ann\+::\+R\+NN}!R\+NN@{R\+NN}} \index{R\+NN@{R\+NN}!mlpack\+::ann\+::\+R\+NN@{mlpack\+::ann\+::\+R\+NN}} \subsubsection{R\+N\+N()} {\footnotesize\ttfamily \textbf{ R\+NN} (\begin{DoxyParamCaption}\item[{const size\+\_\+t}]{rho, }\item[{const bool}]{single = {\ttfamily false}, }\item[{Output\+Layer\+Type}]{output\+Layer = {\ttfamily OutputLayerType()}, }\item[{Initialization\+Rule\+Type}]{initialize\+Rule = {\ttfamily InitializationRuleType()} }\end{DoxyParamCaption})} Create the \doxyref{R\+NN}{p.}{classmlpack_1_1ann_1_1RNN} object. Optionally, specify which initialize rule and performance function should be used. If you want to pass in a parameter and discard the original parameter object, be sure to use std\+::move to avoid unnecessary copy. \begin{DoxyParams}{Parameters} {\em rho} & Maximum number of steps to backpropagate through time (B\+P\+TT). \\ \hline {\em single} & Predict only the last element of the input sequence. \\ \hline {\em output\+Layer} & Output layer used to evaluate the network. \\ \hline {\em initialize\+Rule} & Optional instantiated Initialization\+Rule object for initializing the network parameter. \\ \hline \end{DoxyParams} \mbox{\label{classmlpack_1_1ann_1_1RNN_abbaa486d6d0992486ee3283f1800e20c}} \index{mlpack\+::ann\+::\+R\+NN@{mlpack\+::ann\+::\+R\+NN}!````~R\+NN@{$\sim$\+R\+NN}} \index{````~R\+NN@{$\sim$\+R\+NN}!mlpack\+::ann\+::\+R\+NN@{mlpack\+::ann\+::\+R\+NN}} \subsubsection{$\sim$\+R\+N\+N()} {\footnotesize\ttfamily $\sim$\textbf{ R\+NN} (\begin{DoxyParamCaption}{ }\end{DoxyParamCaption})} Destructor to release allocated memory. \subsection{Member Function Documentation} \mbox{\label{classmlpack_1_1ann_1_1RNN_a8b5234495846c00f6b2c8296ca6bc718}} \index{mlpack\+::ann\+::\+R\+NN@{mlpack\+::ann\+::\+R\+NN}!Add@{Add}} \index{Add@{Add}!mlpack\+::ann\+::\+R\+NN@{mlpack\+::ann\+::\+R\+NN}} \subsubsection{Add()\hspace{0.1cm}{\footnotesize\ttfamily [1/2]}} {\footnotesize\ttfamily void \textbf{ Add} (\begin{DoxyParamCaption}\item[{Args...}]{args }\end{DoxyParamCaption})\hspace{0.3cm}{\ttfamily [inline]}} Definition at line 272 of file rnn.\+hpp. \mbox{\label{classmlpack_1_1ann_1_1RNN_a503a807740e6c729be9efc89520db728}} \index{mlpack\+::ann\+::\+R\+NN@{mlpack\+::ann\+::\+R\+NN}!Add@{Add}} \index{Add@{Add}!mlpack\+::ann\+::\+R\+NN@{mlpack\+::ann\+::\+R\+NN}} \subsubsection{Add()\hspace{0.1cm}{\footnotesize\ttfamily [2/2]}} {\footnotesize\ttfamily void \textbf{ Add} (\begin{DoxyParamCaption}\item[{\textbf{ Layer\+Types}$<$ Custom\+Layers... $>$}]{layer }\end{DoxyParamCaption})\hspace{0.3cm}{\ttfamily [inline]}} Definition at line 279 of file rnn.\+hpp. \mbox{\label{classmlpack_1_1ann_1_1RNN_a3e02a8743fd14b2a902a2e090da2df47}} \index{mlpack\+::ann\+::\+R\+NN@{mlpack\+::ann\+::\+R\+NN}!Evaluate@{Evaluate}} \index{Evaluate@{Evaluate}!mlpack\+::ann\+::\+R\+NN@{mlpack\+::ann\+::\+R\+NN}} \subsubsection{Evaluate()\hspace{0.1cm}{\footnotesize\ttfamily [1/2]}} {\footnotesize\ttfamily double Evaluate (\begin{DoxyParamCaption}\item[{const arma\+::mat \&}]{parameters, }\item[{const size\+\_\+t}]{begin, }\item[{const size\+\_\+t}]{batch\+Size, }\item[{const bool}]{deterministic }\end{DoxyParamCaption})} Evaluate the recurrent neural network with the given parameters. This function is usually called by the optimizer to train the model. \begin{DoxyParams}{Parameters} {\em parameters} & Matrix model parameters. \\ \hline {\em begin} & Index of the starting point to use for objective function evaluation. \\ \hline {\em batch\+Size} & Number of points to be passed at a time to use for objective function evaluation. \\ \hline {\em deterministic} & Whether or not to train or test the model. Note some layer act differently in training or testing mode. \\ \hline \end{DoxyParams} \mbox{\label{classmlpack_1_1ann_1_1RNN_a8a04cfd951b52327d7f2e148c68f365d}} \index{mlpack\+::ann\+::\+R\+NN@{mlpack\+::ann\+::\+R\+NN}!Evaluate@{Evaluate}} \index{Evaluate@{Evaluate}!mlpack\+::ann\+::\+R\+NN@{mlpack\+::ann\+::\+R\+NN}} \subsubsection{Evaluate()\hspace{0.1cm}{\footnotesize\ttfamily [2/2]}} {\footnotesize\ttfamily double Evaluate (\begin{DoxyParamCaption}\item[{const arma\+::mat \&}]{parameters, }\item[{const size\+\_\+t}]{begin, }\item[{const size\+\_\+t}]{batch\+Size }\end{DoxyParamCaption})} Evaluate the recurrent neural network with the given parameters. This function is usually called by the optimizer to train the model. This just calls the other overload of \doxyref{Evaluate()}{p.}{classmlpack_1_1ann_1_1RNN_a3e02a8743fd14b2a902a2e090da2df47} with deterministic = true. \begin{DoxyParams}{Parameters} {\em parameters} & Matrix model parameters. \\ \hline {\em begin} & Index of the starting point to use for objective function evaluation. \\ \hline {\em batch\+Size} & Number of points to be passed at a time to use for objective function evaluation. \\ \hline \end{DoxyParams} \mbox{\label{classmlpack_1_1ann_1_1RNN_a3e01e9e3fe4f5bd8cfc78521567a0f5a}} \index{mlpack\+::ann\+::\+R\+NN@{mlpack\+::ann\+::\+R\+NN}!Evaluate\+With\+Gradient@{Evaluate\+With\+Gradient}} \index{Evaluate\+With\+Gradient@{Evaluate\+With\+Gradient}!mlpack\+::ann\+::\+R\+NN@{mlpack\+::ann\+::\+R\+NN}} \subsubsection{Evaluate\+With\+Gradient()} {\footnotesize\ttfamily double Evaluate\+With\+Gradient (\begin{DoxyParamCaption}\item[{const arma\+::mat \&}]{parameters, }\item[{const size\+\_\+t}]{begin, }\item[{Grad\+Type \&}]{gradient, }\item[{const size\+\_\+t}]{batch\+Size }\end{DoxyParamCaption})} Evaluate the recurrent neural network with the given parameters. This function is usually called by the optimizer to train the model. \begin{DoxyParams}{Parameters} {\em parameters} & Matrix model parameters. \\ \hline {\em begin} & Index of the starting point to use for objective function evaluation. \\ \hline {\em gradient} & Matrix to output gradient into. \\ \hline {\em batch\+Size} & Number of points to be passed at a time to use for objective function evaluation. \\ \hline \end{DoxyParams} \mbox{\label{classmlpack_1_1ann_1_1RNN_aca73798d93d56b280185c01502d8bd13}} \index{mlpack\+::ann\+::\+R\+NN@{mlpack\+::ann\+::\+R\+NN}!Gradient@{Gradient}} \index{Gradient@{Gradient}!mlpack\+::ann\+::\+R\+NN@{mlpack\+::ann\+::\+R\+NN}} \subsubsection{Gradient()} {\footnotesize\ttfamily void Gradient (\begin{DoxyParamCaption}\item[{const arma\+::mat \&}]{parameters, }\item[{const size\+\_\+t}]{begin, }\item[{arma\+::mat \&}]{gradient, }\item[{const size\+\_\+t}]{batch\+Size }\end{DoxyParamCaption})} Evaluate the gradient of the recurrent neural network with the given parameters, and with respect to only one point in the dataset. This is useful for optimizers such as S\+GD, which require a separable objective function. \begin{DoxyParams}{Parameters} {\em parameters} & Matrix of the model parameters to be optimized. \\ \hline {\em begin} & Index of the starting point to use for objective function gradient evaluation. \\ \hline {\em gradient} & Matrix to output gradient into. \\ \hline {\em batch\+Size} & Number of points to be processed as a batch for objective function gradient evaluation. \\ \hline \end{DoxyParams} Referenced by R\+N\+N$<$ Output\+Layer\+Type, Initialization\+Rule\+Type, Custom\+Layers... $>$\+::\+Predictors(). \mbox{\label{classmlpack_1_1ann_1_1RNN_a1fa76af34a6e3ea927b307f0c318ee4b}} \index{mlpack\+::ann\+::\+R\+NN@{mlpack\+::ann\+::\+R\+NN}!Num\+Functions@{Num\+Functions}} \index{Num\+Functions@{Num\+Functions}!mlpack\+::ann\+::\+R\+NN@{mlpack\+::ann\+::\+R\+NN}} \subsubsection{Num\+Functions()} {\footnotesize\ttfamily size\+\_\+t Num\+Functions (\begin{DoxyParamCaption}{ }\end{DoxyParamCaption}) const\hspace{0.3cm}{\ttfamily [inline]}} Return the number of separable functions (the number of predictor points). Definition at line 282 of file rnn.\+hpp. \mbox{\label{classmlpack_1_1ann_1_1RNN_aa68d74dc1e86e4352e00a3cab83a0e4a}} \index{mlpack\+::ann\+::\+R\+NN@{mlpack\+::ann\+::\+R\+NN}!Parameters@{Parameters}} \index{Parameters@{Parameters}!mlpack\+::ann\+::\+R\+NN@{mlpack\+::ann\+::\+R\+NN}} \subsubsection{Parameters()\hspace{0.1cm}{\footnotesize\ttfamily [1/2]}} {\footnotesize\ttfamily const arma\+::mat\& Parameters (\begin{DoxyParamCaption}{ }\end{DoxyParamCaption}) const\hspace{0.3cm}{\ttfamily [inline]}} Return the initial point for the optimization. Definition at line 285 of file rnn.\+hpp. \mbox{\label{classmlpack_1_1ann_1_1RNN_a043f0ccd62e6711a18e0d81047be9a0a}} \index{mlpack\+::ann\+::\+R\+NN@{mlpack\+::ann\+::\+R\+NN}!Parameters@{Parameters}} \index{Parameters@{Parameters}!mlpack\+::ann\+::\+R\+NN@{mlpack\+::ann\+::\+R\+NN}} \subsubsection{Parameters()\hspace{0.1cm}{\footnotesize\ttfamily [2/2]}} {\footnotesize\ttfamily arma\+::mat\& Parameters (\begin{DoxyParamCaption}{ }\end{DoxyParamCaption})\hspace{0.3cm}{\ttfamily [inline]}} Modify the initial point for the optimization. Definition at line 287 of file rnn.\+hpp. \mbox{\label{classmlpack_1_1ann_1_1RNN_a01373d6adb1a306eb2b093c26d5d9031}} \index{mlpack\+::ann\+::\+R\+NN@{mlpack\+::ann\+::\+R\+NN}!Predict@{Predict}} \index{Predict@{Predict}!mlpack\+::ann\+::\+R\+NN@{mlpack\+::ann\+::\+R\+NN}} \subsubsection{Predict()} {\footnotesize\ttfamily void Predict (\begin{DoxyParamCaption}\item[{arma\+::cube}]{predictors, }\item[{arma\+::cube \&}]{results, }\item[{const size\+\_\+t}]{batch\+Size = {\ttfamily 256} }\end{DoxyParamCaption})} Predict the responses to a given set of predictors. The responses will reflect the output of the given output layer as returned by the output layer function. If you want to pass in a parameter and discard the original parameter object, be sure to use std\+::move to avoid unnecessary copy. The format of the data should be as follows\+: \begin{DoxyItemize} \item each slice should correspond to a time step \item each column should correspond to a data point \item each row should correspond to a dimension So, e.\+g., predictors(i, j, k) is the i\textquotesingle{}th dimension of the j\textquotesingle{}th data point at time slice k. The responses will be in the same format. \end{DoxyItemize} \begin{DoxyParams}{Parameters} {\em predictors} & Input predictors. \\ \hline {\em results} & Matrix to put output predictions of responses into. \\ \hline {\em batch\+Size} & Number of points to predict at once. \\ \hline \end{DoxyParams} \mbox{\label{classmlpack_1_1ann_1_1RNN_ae1efcab525131a0bc040aff1a8bb8e2d}} \index{mlpack\+::ann\+::\+R\+NN@{mlpack\+::ann\+::\+R\+NN}!Predictors@{Predictors}} \index{Predictors@{Predictors}!mlpack\+::ann\+::\+R\+NN@{mlpack\+::ann\+::\+R\+NN}} \subsubsection{Predictors()\hspace{0.1cm}{\footnotesize\ttfamily [1/2]}} {\footnotesize\ttfamily const arma\+::cube\& Predictors (\begin{DoxyParamCaption}{ }\end{DoxyParamCaption}) const\hspace{0.3cm}{\ttfamily [inline]}} Get the matrix of data points (predictors). Definition at line 300 of file rnn.\+hpp. \mbox{\label{classmlpack_1_1ann_1_1RNN_a22c21d155cb5637e7d4e821564b27072}} \index{mlpack\+::ann\+::\+R\+NN@{mlpack\+::ann\+::\+R\+NN}!Predictors@{Predictors}} \index{Predictors@{Predictors}!mlpack\+::ann\+::\+R\+NN@{mlpack\+::ann\+::\+R\+NN}} \subsubsection{Predictors()\hspace{0.1cm}{\footnotesize\ttfamily [2/2]}} {\footnotesize\ttfamily arma\+::cube\& Predictors (\begin{DoxyParamCaption}{ }\end{DoxyParamCaption})\hspace{0.3cm}{\ttfamily [inline]}} Modify the matrix of data points (predictors). Definition at line 302 of file rnn.\+hpp. \mbox{\label{classmlpack_1_1ann_1_1RNN_a372de693ad40b3f42839c8ec6ac845f4}} \index{mlpack\+::ann\+::\+R\+NN@{mlpack\+::ann\+::\+R\+NN}!Reset@{Reset}} \index{Reset@{Reset}!mlpack\+::ann\+::\+R\+NN@{mlpack\+::ann\+::\+R\+NN}} \subsubsection{Reset()} {\footnotesize\ttfamily void Reset (\begin{DoxyParamCaption}{ }\end{DoxyParamCaption})} Reset the state of the network. This ensures that all internally-\/held gradients are set to 0, all memory cells are reset, and the parameters matrix is the right size. Referenced by R\+N\+N$<$ Output\+Layer\+Type, Initialization\+Rule\+Type, Custom\+Layers... $>$\+::\+Predictors(). \mbox{\label{classmlpack_1_1ann_1_1RNN_a7178038c3cb8d247eadb94cd2058c432}} \index{mlpack\+::ann\+::\+R\+NN@{mlpack\+::ann\+::\+R\+NN}!Reset\+Parameters@{Reset\+Parameters}} \index{Reset\+Parameters@{Reset\+Parameters}!mlpack\+::ann\+::\+R\+NN@{mlpack\+::ann\+::\+R\+NN}} \subsubsection{Reset\+Parameters()} {\footnotesize\ttfamily void Reset\+Parameters (\begin{DoxyParamCaption}{ }\end{DoxyParamCaption})} Reset the module information (weights/parameters). Referenced by R\+N\+N$<$ Output\+Layer\+Type, Initialization\+Rule\+Type, Custom\+Layers... $>$\+::\+Predictors(). \mbox{\label{classmlpack_1_1ann_1_1RNN_ac109136a291d66f916e382c08a04a8a6}} \index{mlpack\+::ann\+::\+R\+NN@{mlpack\+::ann\+::\+R\+NN}!Responses@{Responses}} \index{Responses@{Responses}!mlpack\+::ann\+::\+R\+NN@{mlpack\+::ann\+::\+R\+NN}} \subsubsection{Responses()\hspace{0.1cm}{\footnotesize\ttfamily [1/2]}} {\footnotesize\ttfamily const arma\+::cube\& Responses (\begin{DoxyParamCaption}{ }\end{DoxyParamCaption}) const\hspace{0.3cm}{\ttfamily [inline]}} Get the matrix of responses to the input data points. Definition at line 295 of file rnn.\+hpp. \mbox{\label{classmlpack_1_1ann_1_1RNN_a1b3ccc9c0245210305d8dd2df8693f3c}} \index{mlpack\+::ann\+::\+R\+NN@{mlpack\+::ann\+::\+R\+NN}!Responses@{Responses}} \index{Responses@{Responses}!mlpack\+::ann\+::\+R\+NN@{mlpack\+::ann\+::\+R\+NN}} \subsubsection{Responses()\hspace{0.1cm}{\footnotesize\ttfamily [2/2]}} {\footnotesize\ttfamily arma\+::cube\& Responses (\begin{DoxyParamCaption}{ }\end{DoxyParamCaption})\hspace{0.3cm}{\ttfamily [inline]}} Modify the matrix of responses to the input data points. Definition at line 297 of file rnn.\+hpp. \mbox{\label{classmlpack_1_1ann_1_1RNN_a8cc9a89f8ec7de47890a86665a393450}} \index{mlpack\+::ann\+::\+R\+NN@{mlpack\+::ann\+::\+R\+NN}!Rho@{Rho}} \index{Rho@{Rho}!mlpack\+::ann\+::\+R\+NN@{mlpack\+::ann\+::\+R\+NN}} \subsubsection{Rho()\hspace{0.1cm}{\footnotesize\ttfamily [1/2]}} {\footnotesize\ttfamily const size\+\_\+t\& Rho (\begin{DoxyParamCaption}{ }\end{DoxyParamCaption}) const\hspace{0.3cm}{\ttfamily [inline]}} Return the maximum length of backpropagation through time. Definition at line 290 of file rnn.\+hpp. \mbox{\label{classmlpack_1_1ann_1_1RNN_aeb617af2894a3e4bbabcd7ebc30a35af}} \index{mlpack\+::ann\+::\+R\+NN@{mlpack\+::ann\+::\+R\+NN}!Rho@{Rho}} \index{Rho@{Rho}!mlpack\+::ann\+::\+R\+NN@{mlpack\+::ann\+::\+R\+NN}} \subsubsection{Rho()\hspace{0.1cm}{\footnotesize\ttfamily [2/2]}} {\footnotesize\ttfamily size\+\_\+t\& Rho (\begin{DoxyParamCaption}{ }\end{DoxyParamCaption})\hspace{0.3cm}{\ttfamily [inline]}} Modify the maximum length of backpropagation through time. Definition at line 292 of file rnn.\+hpp. \mbox{\label{classmlpack_1_1ann_1_1RNN_af0dd9205158ccf7bcfcd8ff81f79c927}} \index{mlpack\+::ann\+::\+R\+NN@{mlpack\+::ann\+::\+R\+NN}!serialize@{serialize}} \index{serialize@{serialize}!mlpack\+::ann\+::\+R\+NN@{mlpack\+::ann\+::\+R\+NN}} \subsubsection{serialize()} {\footnotesize\ttfamily void serialize (\begin{DoxyParamCaption}\item[{Archive \&}]{ar, }\item[{const unsigned}]{int }\end{DoxyParamCaption})} Serialize the model. Referenced by R\+N\+N$<$ Output\+Layer\+Type, Initialization\+Rule\+Type, Custom\+Layers... $>$\+::\+Predictors(). \mbox{\label{classmlpack_1_1ann_1_1RNN_a2697cc8b37d7bca7c055228382a9b208}} \index{mlpack\+::ann\+::\+R\+NN@{mlpack\+::ann\+::\+R\+NN}!Shuffle@{Shuffle}} \index{Shuffle@{Shuffle}!mlpack\+::ann\+::\+R\+NN@{mlpack\+::ann\+::\+R\+NN}} \subsubsection{Shuffle()} {\footnotesize\ttfamily void Shuffle (\begin{DoxyParamCaption}{ }\end{DoxyParamCaption})} Shuffle the order of function visitation. This may be called by the optimizer. \mbox{\label{classmlpack_1_1ann_1_1RNN_a5f18ddeb54971c09eadc5d91dd589892}} \index{mlpack\+::ann\+::\+R\+NN@{mlpack\+::ann\+::\+R\+NN}!Train@{Train}} \index{Train@{Train}!mlpack\+::ann\+::\+R\+NN@{mlpack\+::ann\+::\+R\+NN}} \subsubsection{Train()\hspace{0.1cm}{\footnotesize\ttfamily [1/2]}} {\footnotesize\ttfamily double Train (\begin{DoxyParamCaption}\item[{arma\+::cube}]{predictors, }\item[{arma\+::cube}]{responses, }\item[{Optimizer\+Type \&}]{optimizer, }\item[{Callback\+Types \&\&...}]{callbacks }\end{DoxyParamCaption})} Train the recurrent neural network on the given input data using the given optimizer. This will use the existing model parameters as a starting point for the optimization. If this is not what you want, then you should access the parameters vector directly with \doxyref{Parameters()}{p.}{classmlpack_1_1ann_1_1RNN_a043f0ccd62e6711a18e0d81047be9a0a} and modify it as desired. If you want to pass in a parameter and discard the original parameter object, be sure to use std\+::move to avoid unnecessary copy. The format of the data should be as follows\+: \begin{DoxyItemize} \item each slice should correspond to a time step \item each column should correspond to a data point \item each row should correspond to a dimension So, e.\+g., predictors(i, j, k) is the i\textquotesingle{}th dimension of the j\textquotesingle{}th data point at time slice k. \end{DoxyItemize} \begin{DoxyTemplParams}{Template Parameters} {\em Optimizer\+Type} & Type of optimizer to use to train the model. \\ \hline {\em Callback\+Types} & Types of Callback Functions. \\ \hline \end{DoxyTemplParams} \begin{DoxyParams}{Parameters} {\em predictors} & Input training variables. \\ \hline {\em responses} & Outputs results from input training variables. \\ \hline {\em optimizer} & Instantiated optimizer used to train the model. \\ \hline {\em callbacks} & Callback function for ensmallen optimizer {\ttfamily Optimizer\+Type}. See {\tt https\+://www.\+ensmallen.\+org/docs.\+html\#callback-\/documentation}. \\ \hline \end{DoxyParams} \begin{DoxyReturn}{Returns} The final objective of the trained model (NaN or Inf on error). \end{DoxyReturn} \mbox{\label{classmlpack_1_1ann_1_1RNN_a21493665751d90180fcb7b9e8115111f}} \index{mlpack\+::ann\+::\+R\+NN@{mlpack\+::ann\+::\+R\+NN}!Train@{Train}} \index{Train@{Train}!mlpack\+::ann\+::\+R\+NN@{mlpack\+::ann\+::\+R\+NN}} \subsubsection{Train()\hspace{0.1cm}{\footnotesize\ttfamily [2/2]}} {\footnotesize\ttfamily double Train (\begin{DoxyParamCaption}\item[{arma\+::cube}]{predictors, }\item[{arma\+::cube}]{responses, }\item[{Callback\+Types \&\&...}]{callbacks }\end{DoxyParamCaption})} Train the recurrent neural network on the given input data. By default, the S\+GD optimization algorithm is used, but others can be specified (such as ens\+::\+R\+M\+Sprop). This will use the existing model parameters as a starting point for the optimization. If this is not what you want, then you should access the parameters vector directly with \doxyref{Parameters()}{p.}{classmlpack_1_1ann_1_1RNN_a043f0ccd62e6711a18e0d81047be9a0a} and modify it as desired. If you want to pass in a parameter and discard the original parameter object, be sure to use std\+::move to avoid unnecessary copy. The format of the data should be as follows\+: \begin{DoxyItemize} \item each slice should correspond to a time step \item each column should correspond to a data point \item each row should correspond to a dimension So, e.\+g., predictors(i, j, k) is the i\textquotesingle{}th dimension of the j\textquotesingle{}th data point at time slice k. \end{DoxyItemize} \begin{DoxyTemplParams}{Template Parameters} {\em Optimizer\+Type} & Type of optimizer to use to train the model. \\ \hline {\em Callback\+Types} & Types of Callback Functions. \\ \hline \end{DoxyTemplParams} \begin{DoxyParams}{Parameters} {\em predictors} & Input training variables. \\ \hline {\em responses} & Outputs results from input training variables. \\ \hline {\em callbacks} & Callback function for ensmallen optimizer {\ttfamily Optimizer\+Type}. See {\tt https\+://www.\+ensmallen.\+org/docs.\+html\#callback-\/documentation}. \\ \hline \end{DoxyParams} \begin{DoxyReturn}{Returns} The final objective of the trained model (NaN or Inf on error). \end{DoxyReturn} \mbox{\label{classmlpack_1_1ann_1_1RNN_a5a15e2d0145f9a32c5b439d9d1949908}} \index{mlpack\+::ann\+::\+R\+NN@{mlpack\+::ann\+::\+R\+NN}!Warn\+Message\+Max\+Iterations@{Warn\+Message\+Max\+Iterations}} \index{Warn\+Message\+Max\+Iterations@{Warn\+Message\+Max\+Iterations}!mlpack\+::ann\+::\+R\+NN@{mlpack\+::ann\+::\+R\+NN}} \subsubsection{Warn\+Message\+Max\+Iterations()\hspace{0.1cm}{\footnotesize\ttfamily [1/2]}} {\footnotesize\ttfamily std\+::enable\+\_\+if$<$ Has\+Max\+Iterations$<$Optimizer\+Type, size\+\_\+t\&(Optimizer\+Type\+::$\ast$)()$>$\+::value, void$>$\+::type Warn\+Message\+Max\+Iterations (\begin{DoxyParamCaption}\item[{Optimizer\+Type \&}]{optimizer, }\item[{size\+\_\+t}]{samples }\end{DoxyParamCaption}) const} Check if the optimizer has Max\+Iterations() parameter, if it does then check if it\textquotesingle{}s value is less than the number of datapoints in the dataset. \begin{DoxyTemplParams}{Template Parameters} {\em Optimizer\+Type} & Type of optimizer to use to train the model. \\ \hline \end{DoxyTemplParams} \begin{DoxyParams}{Parameters} {\em optimizer} & optimizer used in the training process. \\ \hline {\em samples} & Number of datapoints in the dataset. \\ \hline \end{DoxyParams} \mbox{\label{classmlpack_1_1ann_1_1RNN_a02d59875a99c37ae210353881619d1af}} \index{mlpack\+::ann\+::\+R\+NN@{mlpack\+::ann\+::\+R\+NN}!Warn\+Message\+Max\+Iterations@{Warn\+Message\+Max\+Iterations}} \index{Warn\+Message\+Max\+Iterations@{Warn\+Message\+Max\+Iterations}!mlpack\+::ann\+::\+R\+NN@{mlpack\+::ann\+::\+R\+NN}} \subsubsection{Warn\+Message\+Max\+Iterations()\hspace{0.1cm}{\footnotesize\ttfamily [2/2]}} {\footnotesize\ttfamily std\+::enable\+\_\+if$<$ !Has\+Max\+Iterations$<$Optimizer\+Type, size\+\_\+t\&(Optimizer\+Type\+::$\ast$)()$>$\+::value, void$>$\+::type Warn\+Message\+Max\+Iterations (\begin{DoxyParamCaption}\item[{Optimizer\+Type \&}]{optimizer, }\item[{size\+\_\+t}]{samples }\end{DoxyParamCaption}) const} Check if the optimizer has Max\+Iterations() parameter, if it doesn\textquotesingle{}t then simply return from the function. \begin{DoxyTemplParams}{Template Parameters} {\em Optimizer\+Type} & Type of optimizer to use to train the model. \\ \hline \end{DoxyTemplParams} \begin{DoxyParams}{Parameters} {\em optimizer} & optimizer used in the training process. \\ \hline {\em samples} & Number of datapoints in the dataset. \\ \hline \end{DoxyParams} The documentation for this class was generated from the following file\+:\begin{DoxyCompactItemize} \item /var/www/mlpack.\+ratml.\+org/mlpack.\+org/\+\_\+src/mlpack-\/3.\+3.\+0/src/mlpack/methods/ann/\textbf{ rnn.\+hpp}\end{DoxyCompactItemize}