Package deepboof.graph

Class FunctionSequence<T extends Tensor<T>,F extends Function<T>>

java.lang.Object

deepboof.graph.FunctionSequence<T,F>

public class FunctionSequence<T extends Tensor<T>,F extends Function<T>>
extends Object

Processes a sequence of forward functions. Any non-cyclical graph with a single input and a single output can be processed by this function. The list of functions passed in to the constructor is assumed to have already been ordered.

Field Summary

Fields

Modifier and Type	Field	Description
protected TensorFactory<T>	factory
protected Map<String,Node<T,F>>	lookup
protected Map<String,Tuple2<T,T>>	outputStorage
protected List<Node<T,F>>	sequence

Constructor Summary

Constructors

Constructor	Description
FunctionSequence(List<Node<T,F>> sequence, Class<T> type)	Configures the sequence

Method Summary

Modifier and Type	Method	Description
T	getNodeOutput(int index)
int[]	getOutputShape()
List<Node<T,F>>	getSequence()
Class<T>	getTensorType()
void	initialize(int[] inputShape)	Initialize and declare memory for all nodes in the network given the shape of the input tensor
void	process(T input, T output)	Process the input tensor and compute the output tensor by feeding the results through the network.
void	setParameters(Map<String,List<T>> nodeParameters)	Specify the parameters for each node in the network

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Field Details

sequence

protected List<Node<T extends Tensor<T>,F extends Function<T>>> sequence

lookup

protected Map<String,Node<T extends Tensor<T>,F extends Function<T>>> lookup

outputStorage

protected Map<String,Tuple2<T extends Tensor<T>,T extends Tensor<T>>> outputStorage

factory

protected TensorFactory<T extends Tensor<T>> factory

Constructor Details

FunctionSequence

public FunctionSequence(List<Node<T,F>> sequence,
Class<T> type)

Configures the sequence

Parameters:

sequence - Sequence of functions which has been ordered to meet pre-requisites.

type - Type of tensor

Method Details

initialize

public void initialize(int[] inputShape)

Initialize and declare memory for all nodes in the network given the shape of the input tensor

Parameters:

inputShape - Shape of input tensor.

setParameters

public void setParameters(Map<String,List<T>> nodeParameters)

Specify the parameters for each node in the network

Parameters:

nodeParameters - Map where the key is the function/node name and the value is the parameters for that node

process

public void process(T input,
T output)

Process the input tensor and compute the output tensor by feeding the results through the network. Must call initialize(int[]) once with the same shape as the input tensor. Must also call setParameters(java.util.Map<java.lang.String, java.util.List<T>>)

Parameters:

input - Input tensor

output - Storage for output tensor.

getSequence

public List<Node<T,F>> getSequence()

getNodeOutput

public T getNodeOutput(int index)

getOutputShape

public int[] getOutputShape()

getTensorType

public Class<T> getTensorType()