org.simbrain.network.util

Class SimnetUtils

java.lang.Object

org.simbrain.network.util.SimnetUtils

public class SimnetUtils
extends java.lang.Object

SimnetUtils provides utility classes relating to Simbrain networks.

Author:

jyoshimi, ztosi

Constructor Summary

Constructors

Constructor and Description
SimnetUtils()

Method Summary

Modifier and Type	Method and Description
static double	findMaxEig(double[][] weightMatrix) Find the largest eigenvalue for the provided matrix.
static java.util.List<java.util.List<Neuron>>	getIntermedateLayers(Network network, java.util.List<Neuron> sourceLayer, java.util.List<Neuron> targetLayer) Given a source and target set of neurons, find all layers of neurons connecting them, as follows.
static java.awt.geom.Point2D	getUpperLeft(java.util.List<java.lang.Object> objects) Return the upper left corner of a list of objects, based on neurons.
static Synapse[][]	getWeightMatrix(java.util.List<Neuron> srcLayer, java.util.List<Neuron> targetLayer) Gets a matrix of Synapse objects, formatted like the getWeights method.
static double[][]	getWeights(java.util.List<Neuron> srcLayer, java.util.List<Neuron> targetLayer) Returns the weights connecting two lists of neurons as an N x M matrix of doubles, where N is the number of source neurons, and M is the number of target neurons.
static void	printLayers(java.util.List<java.util.List<Neuron>> layers) Prints a group of layers (a list of lists of neurons), typically for debugging.
static void	scaleEigenvalue(java.util.List<Neuron> src, java.util.List<Neuron> tar, double desiredEigen)
static void	scaleWeights(java.util.List<Neuron> src, java.util.List<Neuron> tar, double scalar) Scales weights connecting source and target lists.
static void	setWeights(java.util.List<Neuron> src, java.util.List<Neuron> tar, double[][] w) Set the weights connecting two lists of neurons using a weight matrix.

Methods inherited from class java.lang.Object

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

Constructor Detail

SimnetUtils

public SimnetUtils()

Method Detail

getWeights

public static double[][] getWeights(java.util.List<Neuron> srcLayer,
                                    java.util.List<Neuron> targetLayer)

Returns the weights connecting two lists of neurons as an N x M matrix of doubles, where N is the number of source neurons, and M is the number of target neurons. That is, each row of the matrix corresponds to a source neuron's fan-out weight vector.

Parameters:

srcLayer - source layer

targetLayer - target layer

Returns:

weight matrix

setWeights

public static void setWeights(java.util.List<Neuron> src,
                              java.util.List<Neuron> tar,
                              double[][] w)

Set the weights connecting two lists of neurons using a weight matrix. Assumes that each row of the matrix corresponds to a source neuron's fan-out weight vector, as above. If a weight is missing it is added to the root network (from where it can in some cases be routed to a SynapseGroup)

Parameters:

src - the list of source neurons

tar - the list of target neurons

w - the new weight values for the network.

getWeightMatrix

public static Synapse[][] getWeightMatrix(java.util.List<Neuron> srcLayer,
                                          java.util.List<Neuron> targetLayer)

Gets a matrix of Synapse objects, formatted like the getWeights method. Non-existence synapses are given a null value.

Parameters:

srcLayer - source neurons

targetLayer - target neurons

Returns:

the matrix of synapses.

scaleWeights

public static void scaleWeights(java.util.List<Neuron> src,
                                java.util.List<Neuron> tar,
                                double scalar)

Scales weights connecting source and target lists.

Parameters:

src - source neurons

tar - target neurons

scalar - scalar value which is multiplied by the weight matrix

findMaxEig

public static double findMaxEig(double[][] weightMatrix)

Find the largest eigenvalue for the provided matrix.

Parameters:

weightMatrix - a matrix representation of the weights for use in linear algebraic operations

Returns:

the largest eigenvalue of this matrix by absolute value

scaleEigenvalue

public static void scaleEigenvalue(java.util.List<Neuron> src,
                                   java.util.List<Neuron> tar,
                                   double desiredEigen)

Parameters:

src - list of source neurons

tar - list of target neurons

desiredEigen - : the new max eig or spectral radius for the weight matrix

getUpperLeft

public static java.awt.geom.Point2D getUpperLeft(java.util.List<java.lang.Object> objects)

Return the upper left corner of a list of objects, based on neurons.

Parameters:

objects - list of objects

Returns:

the point corresponding to the upper left corner of the objects

getIntermedateLayers

public static java.util.List<java.util.List<Neuron>> getIntermedateLayers(Network network,
                                                                          java.util.List<Neuron> sourceLayer,
                                                                          java.util.List<Neuron> targetLayer)

Given a source and target set of neurons, find all layers of neurons connecting them, as follows. Assumes a sequence of layers from source to target, each fully connected to the next, and no other connections (e.g. recurrent connections). If a path from the source to target layer is not found then a list containing only the source and target layers is returned.

Parameters:

network - the neural network

sourceLayer - the source neurons

targetLayer - the target neurons

Returns:

the resulting list of layers

printLayers

public static void printLayers(java.util.List<java.util.List<Neuron>> layers)

Prints a group of layers (a list of lists of neurons), typically for debugging.

Parameters:

layers - the layers to print