public override double Forward(Matrix Actual, Matrix Expected, MatrixData data, int layerCount)
{
double error = 0.0;
if (Actual.rows != Expected.rows || Actual.cols != Expected.cols)
{
throw new MatrixException("Actual does not have the same size as Expected");
}
double regularizationValue = 0.0;
for (int i = 0; i < layerCount; i++)
{
regularizationValue = RegularizationFunction.CalculateNorm(data.Data["W" + i.ToString()]);
}
for (int i = 0; i < Actual.rows; i++)
{
for (int j = 0; j < Actual.cols; j++)
{
error += Expected[i, j] * Math.Log(Expected[i, j] / Actual[i, j]);
}
}
error += regularizationValue;
BatchCost += error;
return(error);
}
/// <summary>Constructs a link in the neural network initialized with random weight.</summary>
/// <param name="source">source The source node.</param>
/// <param name="dest">@param dest The destination node.</param>
/// <param name="regularization">regularization The regularization function that computes the penalty for this weight.If null, there will be no regularization.</param>
/// <param name="initZero"></param>
public Link(Node source, Node dest, RegularizationFunction regularization, bool initZero = false)
{
this.id = source.Id + "-" + dest.Id;
this.source = source;
this.dest = dest;
this.regularization = regularization;
if (initZero)
{
this.weight = 0.0;
}
}
public override Matrix Forward(Matrix inputs)
{
Params["W"].InMap(MutationFunction.Mutate);
// Calculate Regularization Value On W and B
RegularizationValue = (float)RegularizationFunction.CalculateNorm(Params["W"]);
// Calculate Feed Forward Operation
Params["X"] = inputs;
Params["Z"] = Params["W"] * Params["X"] + Params["B"];
Params["A"] = ActivationFunction.Forward(Params["Z"]);
return(Params["A"]);
}
public override Matrix Backward(Matrix Actual, Matrix Expected, MatrixData data, int layerCount)
{
double error = 0.0;
if (Actual.rows != Expected.rows || Actual.cols != Expected.cols)
{
throw new MatrixException("Actual Matrix does not have the same size as The Expected Matrix");
}
double regularizationValue = 0.0;
for (int i = 0; i < layerCount; i++)
{
regularizationValue = RegularizationFunction.CalculateNorm(data.Data["W" + i.ToString()]);
}
for (int i = 0; i < Actual.rows; i++)
{
for (int j = 0; j < Actual.cols; j++)
{
error += Math.Pow((Actual[i, j] - Expected[i, j]), 2);
}
}
error /= Tao;
error = Math.Exp(error);
error *= Tao;
Matrix gradMatrix = Actual.Duplicate();
for (int i = 0; i < Actual.rows; i++)
{
for (int j = 0; j < Actual.cols; j++)
{
gradMatrix[i, j] = (Actual[i, j] - Expected[i, j]) * error;
}
}
return(gradMatrix);
}
/* Copyright 2016 Google Inc. All Rights Reserved.
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
* http://www.apache.org/licenses/LICENSE-2.0
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*
* Credits
* This was created by Daniel Smilkov and Shan Carter. This is a continuation
* of many people’s previous work — most notably Andrej Karpathy’s convnet.js
* demo and Chris Olah’s articles about neural networks. Many thanks also to
* D. Sculley for help with the original idea and to Fernanda Viégas and
* Martin Wattenberg and the rest of the Big Picture and Google Brain teams
* for feedback and guidance.
* ==============================================================================*/
/// <summary>Builds a neural network.</summary>
/// <param name="networkShape">The shape of the network. E.g. [1, 2, 3, 1] means the network will have one input node, 2 nodes in first hidden layer, 3 nodes in second hidden layer and 1 output node.</param>
/// <param name="activation">The activation function of every hidden node.</param>
/// <param name="outputActivation">The activation function for the output nodes.</param>
/// <param name="regularization">The regularization function that computes a penalty for a given weight (parameter) in the network. If null, there will be no regularization.</param>
/// <param name="inputIds">List of ids for the input nodes.</param>
public List <List <Node> > BuildNetwork(List <int> networkShape, ActivationFunction activation, ActivationFunction outputActivation, RegularizationFunction regularization, List <string> inputIds, bool initZero = false)
{
int numLayers = networkShape.Count;
int id = 1;
/* List of layers, with each layer being a list of nodes. */
List <List <Node> > network = new List <List <Node> >();
for (int layerIdx = 0; layerIdx < numLayers; layerIdx++)
{
bool isOutputLayer = layerIdx == numLayers - 1;
bool isInputLayer = layerIdx == 0;
List <Node> currentLayer = new List <Node>();
Network.Add(currentLayer);
int numNodes = networkShape[layerIdx];
for (int i = 0; i < numNodes; i++)
{
string nodeId = id.ToString();
if (isInputLayer)
{
nodeId = inputIds[i];
}
else
{
id++;
}
ActivationFunction act;
if (isOutputLayer)
{
act = outputActivation;
}
else
{
act = activation;
}
Node node = new Node(nodeId, act, initZero);
currentLayer.Add(node);
if (layerIdx >= 1)
{
// Add links from nodes in the previous layer to this node.
for (int j = 0; j < network[layerIdx - 1].Count; j++)
{
Node prevNode = network[layerIdx - 1][j];
Link link = new Link(prevNode, node, regularization, initZero);
prevNode.Outputs.Add(link);
node.InputLinks.Add(link);
}
}
}
}
return(network);
}//good
public static List <List <Node> > buildNetwork(List <int> networkShape, ActivationFunction activation,
ActivationFunction outputActivation, RegularizationFunction regularization,
List <string> inputIds, bool initZero)
{
var numLayers = networkShape.Count;
var id = 1;
List <List <Node> > network = new List <List <Node> >();
for (var layerIdx = 0; layerIdx < numLayers; layerIdx++)
{
var isOutputLayer = layerIdx == numLayers - 1;
var isInputLayer = layerIdx == 0;
List <Node> currentLayer = new List <Node>();
network.Add(currentLayer);
int numNodes = networkShape[layerIdx];
for (int i = 0; i < numNodes; i++)
{
var nodeId = id.ToString();
if (isInputLayer)
{
nodeId = inputIds[i];
}
else
{
id++;
}
var node = new Node(nodeId, isOutputLayer ? outputActivation : activation, initZero);
currentLayer.Add(node);
if (layerIdx >= 1)
{
for (int j = 0; j < network[layerIdx - 1].Count; j++)
{
var prevNode = network[layerIdx - 1][j];
var link = new Link(prevNode, node, regularization, initZero);
prevNode.outputs.Add(link);
node.inputLinks.Add(link);
}
}
}
}
return(network);
}