public static NeuralNetwork SimpleLinearNetwork()
{
var net = new NeuralNetwork(1, activation: new Mocks.LinearActivation());
net.IsTraining = true;
var layer1 = new NeuralLayer(1);
net.AddLayer(layer1);
var layer2 = new NeuralLayer(1);
net.AddLayer(layer2);
var layer3 = new NeuralLayer(1);
net.AddLayer(layer3);
net.Build();
layer1[0].Bias = 1;
layer1[0][0] = 3;
layer2[0].Bias = -1;
layer2[0][0] = 1;
layer3[0].Bias = 2;
layer3[0][0] = -1;
return(net);
}
public static NeuralNetwork CreateElmanNN(int inputCount, int outputCount)
{
NeuralNetwork network = new NeuralNetwork();
Layer input = new Layer();
Layer output = new Layer();
for (int i = 0; i < inputCount; i++)
{
input.AddUnit(new NetworkUnit(ActFuncs.Identity));
}
for (int i = 0; i < outputCount; i++)
{
NetworkUnit memoryUnit = new NetworkUnit(ActFuncs.Identity);
NetworkUnit biasUnit = NetworkUnit.CreateBias();
NetworkUnit outUnit = new NetworkUnit(ActFuncs.Sigmoid, biasUnit)
{
MemoryUnit = memoryUnit
};
input.AddUnit(memoryUnit);
output.AddUnit(outUnit);
}
network.AddLayer(input);
network.AddLayer(output);
return(network);
}
void Start()
{
snakeInstances = new List <SnakeInstance>();
//First iteration, 1 network will be unchanged, the rest will have adjusted weights
int x;
int y;
for (int i = 0; i < 88; i++)
{
NeuralNetwork network = new NeuralNetwork();
network.AddLayer(new NeuralLayer(6, 0.5, "INPUT"));
network.AddLayer(new NeuralLayer(100, 0.5, ""));
network.AddLayer(new NeuralLayer(3, 0.5, "OUTPUT"));
network.BuildNetwork();
x = -100 + ((i % 11) * 20);
y = -70 + (Mathf.FloorToInt(i / 11) * 20);
snakeInstances.Add(Instantiate(snakePrefab, new Vector2(x, y), new Quaternion()).GetComponent <SnakeInstance>());
snakeInstances[i].network = network;
if (i > 0)
{
snakeInstances[i].network.RandomizeWeights(0.4);
}
}
//Now that we created all of our instances, start iteration 1
for (int i = 0; i < 88; i++)
{
snakeInstances[i].Begin();
}
}
public static NeuralNetwork SimpleLinearNetworkWithDropout(double drate, double dseed)
{
var net = new NeuralNetwork(1, activation: new Mocks.LinearActivation());
net.IsTraining = true;
var layer1 = new NeuralLayer(1);
net.AddLayer(layer1);
var layer2 = new NeuralLayer(1);
layer2.DropoutRate = drate;
layer2.DropoutSeed = 1;
layer2[0].LastRetained = true;
net.AddLayer(layer2);
var layer3 = new NeuralLayer(1);
net.AddLayer(layer3);
net.Build();
layer1[0].Bias = 1;
layer1[0][0] = 3;
layer2[0].Bias = -1;
layer2[0][0] = 1;
layer3[0].Bias = 2;
layer3[0][0] = -1;
return(net);
}
override public void Init(NeuralNetwork p_network = null)
{
NeuralNetwork network = null;
Optimizer optimizer = null;
if (p_network == null)
{
network = new NeuralNetwork();
network.AddLayer("input", new InputLayer(GetParam(STATE_DIM)), BaseLayer.TYPE.INPUT);
network.AddLayer("hidden0", new CoreLayer(SolverConfig.GetInstance().hidden_layer, ACTIVATION.RELU, BaseLayer.TYPE.HIDDEN), BaseLayer.TYPE.HIDDEN);
network.AddLayer("output", new CoreLayer(GetParam(ACTION_DIM), ACTIVATION.TANH, BaseLayer.TYPE.OUTPUT), BaseLayer.TYPE.OUTPUT);
// feed-forward connections
network.AddConnection("input", "hidden0", Connection.INIT.GLOROT_UNIFORM);
network.AddConnection("hidden0", "output", Connection.INIT.GLOROT_UNIFORM);
}
else
{
network = p_network;
}
optimizer = new ADAM(network);
//optimizer = new RMSProp(network);
//optimizer = new BackProp(network, 1e-5f, 0.99f, true);
_critic = new DeepQLearning(optimizer, network, 0.99f, SolverConfig.GetInstance().memory_size, SolverConfig.GetInstance().batch_size, SolverConfig.GetInstance().qtupdate_size);
_critic.SetAlpha(SolverConfig.GetInstance().learning_rate);
}
private void button1_Click(object sender, EventArgs e)
{
network = new NeuralNetwork();
network.AddLayer(9);
network.AddLayer(7);
button2.Enabled = true;
button3.Enabled = true;
}
public void TestUninitializedNetworkThrowsInvalidOperationExceptionOnCompute()
{
NeuralNetwork nn = new NeuralNetwork();
nn.AddLayer(1, new LinearActivation());
nn.AddLayer(1, new UnipolarSigmoidActivation());
Vector <double> result = nn.ComputeOutput(new double[] { 0 });
}
public void TestNeuralNetworkIsCreatedWith2Layers()
{
NeuralNetwork nn = new NeuralNetwork();
nn.AddLayer(2, new LinearActivation());
nn.AddLayer(1, new UnipolarSigmoidActivation());
nn.Initialize(CreationModes.RandomizeWeights);
Assert.AreEqual(2, nn.LayerCount);
}
public void TestIncorrectInputLengthThrowsInvalidOperationExceptionOnCompute()
{
NeuralNetwork nn = new NeuralNetwork();
nn.AddLayer(1, new LinearActivation());
nn.AddLayer(1, new UnipolarSigmoidActivation());
nn.Initialize(CreationModes.RandomizeWeights);
Vector <double> result = nn.ComputeOutput(new double[] { 0, 1 });
}
public void CreateNetwork()
{
Network = new NeuralNetwork();
Network.AddLayer(InputLayerSize);
Network.AddLayer(HiddenLayerSize);
Network.AddLayer(OutputLayerSize);
Network.LogMessage = Log;
Log?.Invoke("Network Created");
}
public void CreatNeuralNetwork(float lr)
{
nn = new NeuralNetwork(lr, new LossPowSum());
nn.AddLayer(new LinerLayer(2, 3));
nn.AddLayer(new LinerLayer(3, 2));
nn.AddLayer(new LinerLayer(2, 2));
nn.RandomWeigths();
Lr_txt.text = "学习步长：" + lr.ToString() + "f";
Train_Num_txt.text = "训练总次：" + TrainNum.ToString();
Speed_txt.text = "单次时长：" + trainspeed.ToString() + "s";
}
public void TestSeveralNeuronsPerceptronProcessingIndividualWeightModifications()
{
NeuralNetwork nn = new NeuralNetwork();
nn.AddLayer(3, new LinearActivation());
nn.AddLayer(2, new LinearActivation());
nn.Initialize(CreationModes.RandomizeWeights);
nn.SetIncomingWeightsForNeuron(1, 0, new double[] { 1, 0, 2 });
nn.SetIncomingWeightsForNeuron(1, 1, new double[] { 0, 4, 1 });
Vector <double> result = nn.ComputeOutput(new double[] { 1, 2, 3 });
Assert.AreEqual(7.0, result[0], EPSILON);
Assert.AreEqual(11.0, result[1], EPSILON);
}
private void Awake()
{
Func <double, double> sigmoid = (x) => (1f / (1f + Math.Pow(Math.E, -x)));
Func <double, double> linear = (x) => x;
Func <double, double, double> derSigmoid = (x, y) => (y * (1f - y));
Func <double, double, double> derLinear = (x, y) => 1f;
perceptronModel = new NeuralNetwork(63, 1, 0.01f, linear, derLinear);
//model masukin dari belakang :v
backpropModel = new NeuralNetwork(63, 1, 0.01f, sigmoid, derSigmoid);
backpropModel.AddLayer(2, "relu");
backpropModel.AddLayer(4, "relu");
backpropModel.AddLayer(8, "relu");
}
/// <summary>
/// Creates fully connected NN
/// </summary>
/// <param name="topology">Network topology from input to output layer
/// (i.e. [2,10,3] means NN with 2D input, 1 hidden layer with 10 neurons and 3D output)</param>
public static NeuralNetwork CreateFullyConnectedNetwork(int[] topology,
IActivationFunction activation = null,
bool randomizeInitialWeights = true,
int randomSeed = 0)
{
if (topology == null || topology.Length < 2)
{
throw new MLException("Network topology must have at least input and output dimensions");
}
var net = new NeuralNetwork(topology[0], activation);
var lcount = topology.Length - 1;
for (int i = 1; i <= lcount; i++)
{
var neuronCount = topology[i];
var layer = new NeuralLayer(neuronCount);
net.AddLayer(layer);
}
net.Build();
if (randomizeInitialWeights)
{
net.RandomizeParameters(randomSeed);
}
return(net);
}
private NeuralNetwork CreateNewNetwork()
{
//先创建一个神经网络
var network = new NeuralNetwork();
//创建一个全连接层，输出形状为(1, 10, 1)，激活函数为Sigmoid
//var d1 = new DenseLayer(new Shape(1, 10, 1), ActivationType.Sigmoid);
//network.AddLayer(d1);
//创建一个全连接层，输出形状为(1, 2, 1)，注意这里是最后一层，输入形状要和你需要的输出对应，激活函数为Sigmoid
var d2 = new DenseLayer(new Shape(1, 2, 1), ActivationType.Sigmoid);
network.AddLayer(d2);
//初始化参数
var initArgs = new NetworkInitializeArgs();
//输入形状 你的输入数据形状，这里是小车的demo，输入为射线的数量
initArgs.inputShape = new Shape(1, rayNum, 1);
//权重的初始化范围 weight = Random.Range(-0.1f, 0.1f)
initArgs.initWeightRange = (-0.1f, 0.1f);
//偏执项的初始化范围 bias = Random.Range(-0.1f, 0.1f)
initArgs.initBiasRange = (-0.1f, 0.1f);
//初始化
network.Initialize(initArgs);
return(network);
}
public void TestSimpleProcessing3()
{
NeuralNetwork nn = new NeuralNetwork();
nn.AddLayer(1, new LinearActivation());
nn.AddLayer(1, new UnipolarSigmoidActivation());
nn.Initialize(CreationModes.NoAction);
nn.SetIncomingWeightsForLayer(1, new List <double[]>()
{
new double[] { 2 }
});
Vector <double> result = nn.ComputeOutput(new double[] { 0 });
TestContext.WriteLine(result[0].ToString());
Assert.AreEqual(0.5, result[0]);
}
// Start is called before the first frame update
void Start()
{
var network = new NeuralNetwork();
var d1 = new DenseLayer(new Shape(1, 10, 1), ActivationType.Sigmoid);
network.AddLayer(d1);
var initArgs = new NetworkInitializeArgs();
initArgs.inputShape = new Shape(1, 28 * 28, 1);
initArgs.initWeightRange = (-0.1f, 0.1f);
initArgs.initBiasRange = (-0.1f, 0.1f);
network.Initialize(initArgs);
var trainArgs = new NeuralNetworkTrainArgs();
trainArgs.trainingData = ReadTrainingData(); //设置数据
trainArgs.trainingLabels = ReadTrainingLabels(); //设置标签
trainArgs.learningRate = 0.01f; //设置学习速率，越大学习的速度越快，但出现不收敛的可能性也越大
trainArgs.onOnceEpoch = (i) =>
{
var accuracy = GetAccuracy(network, trainArgs.trainingData, trainArgs.trainingLabels);
Debug.Log($"第{i}个训练回合, 准确率:{accuracy}");
};
trainArgs.trainEpoches = 100; //设置训练的回合数
network.Train(trainArgs); //开始训练
TestNetwork(network, trainArgs.trainingData, trainArgs.trainingLabels);
}
public void TestSeveralNeuronsPerceptronProcessing()
{
NeuralNetwork nn = new NeuralNetwork();
nn.AddLayer(3, new LinearActivation());
nn.AddLayer(2, new LinearActivation());
nn.Initialize(CreationModes.NoAction);
nn.SetIncomingWeightsForLayer(1, new List <double[]>()
{
new double[] { 1, 0, 2 }, new double[] { 0, 4, 1 }
});
Vector <double> result = nn.ComputeOutput(new double[] { 1, 2, 3 });
Assert.AreEqual(7.0, result[0], EPSILON);
Assert.AreEqual(11.0, result[1], EPSILON);
}
public void TestNeuralNetworkAttributes5Layers()
{
NeuralNetwork nn = new NeuralNetwork();
nn.AddLayer(2, new LinearActivation());
nn.AddLayer(4, new UnipolarSigmoidActivation());
nn.AddLayer(7, new UnipolarSigmoidActivation());
nn.AddLayer(4, new UnipolarSigmoidActivation());
nn.AddLayer(1, new UnipolarSigmoidActivation());
nn.Initialize(CreationModes.RandomizeWeights);
Assert.AreEqual(5, nn.LayerCount);
Assert.AreEqual(3, nn.HiddenLayerCount);
Assert.AreEqual(2, nn.InputCount);
Assert.AreEqual(1, nn.OutputCount);
Assert.AreEqual(4, nn.LayerAt(3).NeuronCount);
}
public void TestSettingNeuronWeights()
{
NeuralNetwork nn = new NeuralNetwork();
nn.AddLayer(2, new LinearActivation(), false);
nn.AddLayer(3, new LinearActivation(), false);
nn.AddLayer(1, new LinearActivation(), false);
nn.Initialize(CreationModes.NoAction);
nn.SetIncomingWeightsForNeuron(1, 0, new double[] { 1, -2 });
nn.SetIncomingWeightsForNeuron(1, 1, new double[] { 4, 0 });
nn.SetIncomingWeightsForNeuron(1, 2, new double[] { -1, 3 });
nn.SetIncomingWeightsForNeuron(2, 0, new double[] { -1, -2, 3 });
Vector <double> result = nn.ComputeOutput(new double[] { 1.0, 2.0 });
Assert.AreEqual(10.0, result[0], 0.0001);
}
public void TestSimpleMLPProcessing1()
{
NeuralNetwork nn = new NeuralNetwork();
nn.AddLayer(1, new LinearActivation());
nn.AddLayer(1, new LinearActivation());
nn.AddLayer(1, new LinearActivation());
nn.Initialize(CreationModes.NoAction);
nn.SetIncomingWeightsForLayer(1, new List <double[]>()
{
new double[] { 2 }
});
nn.SetIncomingWeightsForLayer(2, new List <double[]>()
{
new double[] { 2 }
});
Vector <double> result = nn.ComputeOutput(new double[] { 3 });
Assert.AreEqual(12, result[0]);
}
public void TestLastOutputInNetworkContext()
{
NeuralNetwork nn = new NeuralNetwork();
nn.AddLayer(2, new LinearActivation(), false);
nn.AddLayer(3, new LinearActivation(), false);
nn.AddLayer(1, new LinearActivation(), false);
nn.Initialize(CreationModes.NoAction);
nn.SetIncomingWeightsForLayer(1, new List <double[]>()
{
new double[] { 1, -2 }, new double[] { 4, 0 }, new double[] { -1, 3 }
});
nn.SetIncomingWeightsForLayer(2, new List <double[]>()
{
new double[] { -1, -2, 3 }
});
Vector <double> result = nn.ComputeOutput(new double[] { 1.0, 2.0 });
double[] lastOutput = nn.LayerAt(1).LastOutput.ToArray();
// TODO: write assert
}
public void TestComplexMLPProcessingWithBiasBipolarSigmoid1()
{
NeuralNetwork nn = new NeuralNetwork();
nn.AddLayer(2, new LinearActivation(), false);
nn.AddLayer(3, new BipolarTanhActivation(), true);
nn.AddLayer(1, new BipolarTanhActivation(), true);
nn.Initialize(CreationModes.NoAction);
nn.SetIncomingWeightsForLayer(1, new List <double[]>()
{
new double[] { 2, 1 }, new double[] { 3, -2 }, new double[] { -2, 7 }
});
nn.SetIncomingWeightsForLayer(2, new List <double[]>()
{
new double[] { 17, 13, 20 }
});
nn.SetBiasForLayer(1, new double[] { -100.0, -100.0, -100.0 });
nn.SetBiasForLayer(2, new double[] { 50.0 });
Vector <double> result = nn.ComputeOutput(new double[] { 0.8, 0.5 });
Assert.AreEqual(0.0, result[0], 0.0001);
}
public void TestSimpleDeepNetProcessing1()
{
NeuralNetwork nn = new NeuralNetwork();
nn.AddLayer(1, new LinearActivation());
nn.AddLayer(1, new LinearActivation());
nn.AddLayer(1, new LinearActivation());
nn.AddLayer(1, new LinearActivation());
nn.AddLayer(1, new LinearActivation());
nn.Initialize(CreationModes.NoAction);
for (int i = 1; i < 5; i++)
{
nn.SetIncomingWeightsForLayer(i, new List <double[]>()
{
new double[] { 2 }
});
}
Vector <double> result = nn.ComputeOutput(new double[] { 3 });
TestContext.WriteLine(result[0].ToString());
Assert.AreEqual(48, result[0]);
}
static void Main(string[] args)
{
var network = new NeuralNetwork(3);
var layerFactory = new NeuralLayerFactory();
network.AddLayer(layerFactory.CreateNeuralLayer(3, new RectifiedActivationFuncion(), new WeightedSumFunction()));
//network.AddLayer(layerFactory.CreateNeuralLayer(3, new RectifiedActivationFuncion(), new WeightedSumFunction()));
network.AddLayer(layerFactory.CreateNeuralLayer(1, new SigmoidActivationFunction(0.7), new WeightedSumFunction()));
/* network.PushExpectedValues(
* new double[][] {
* new double[] { 0 },
* new double[] { 1 },
* new double[] { 1 },
* new double[] { 0 },
* new double[] { 1 },
* new double[] { 0 },
* new double[] { 0 },
* });
*
* network.Train(
* new double[][] {
* new double[] { 150, 2, 0 },
* new double[] { 1002, 56, 1 },
* new double[] { 1060, 59, 1 },
* new double[] { 200, 3, 0 },
* new double[] { 300, 3, 1 },
* new double[] { 120, 1, 0 },
* new double[] { 80, 1, 0 },
* }, 10000);
*
* network.PushInputValues(new double[] { 1054, 54, 1 });
* var outputs = network.GetOutput(); */
}
private static NeuralNetwork InitializeDefaultNeuralNetwork(DataProvider dp)
{
using (CNNDataSet.TrainingRatesDataTable table = new CNNDataSet.TrainingRatesDataTable())
{
table.ReadXml(@"D:\prj\cnnwb\CNNWB.Data\TrainingSchemes\LeCun2.scheme-xml");
CNNDataSet.TrainingRatesRow row = table.Rows[0] as CNNDataSet.TrainingRatesRow;
_data = new TrainingRate(row.Rate, row.Epochs, row.MinimumRate, row.WeightDecayFactor, row.Momentum, row.BatchSize, row.InitialAvgLoss, row.DecayFactor, row.DecayAfterEpochs, row.WeightSaveTreshold, row.Distorted, row.DistortionPercentage, row.SeverityFactor, row.MaxScaling, row.MaxRotation, row.ElasticSigma, row.ElasticScaling);
}
//NeuralNetwork network = new NeuralNetwork(_dp, "LeNet-5", 10, 0.8D, LossFunctions.MeanSquareError,
// DataProviderSets.MNIST, TrainingStrategy.SGDLevenbergMarquardt, 0.02D);
//network.AddLayer(LayerTypes.Input, 1, 32, 32);
//network.AddLayer(LayerTypes.Convolutional, ActivationFunctions.Tanh, 6, 28, 28, 5, 5);
//network.AddLayer(LayerTypes.AvgPooling, ActivationFunctions.Tanh, 6, 14, 14, 2, 2);
//bool[] maps = new bool[6 * 16]
//{
// true, false,false,false,true, true, true, false,false,true, true, true, true, false,true, true,
// true, true, false,false,false,true, true, true, false,false,true, true, true, true, false,true,
// true, true, true, false,false,false,true, true, true, false,false,true, false,true, true, true,
// false,true, true, true, false,false,true, true, true, true, false,false,true, false,true, true,
// false,false,true, true, true, false,false,true, true, true, true, false,true, true, false,true,
// false,false,false,true, true, true, false,false,true, true, true, true, false,true, true, true
//};
//network.AddLayer(LayerTypes.Convolutional, ActivationFunctions.Tanh, 16, 10, 10, 5, 5, mappings: new Mappings(maps));
//network.AddLayer(LayerTypes.AvgPooling, ActivationFunctions.Tanh, 16, 5, 5, 2, 2);
//network.AddLayer(LayerTypes.Convolutional, ActivationFunctions.Tanh, 120, 1, 1, 5, 5);
//network.AddLayer(LayerTypes.FullyConnected, ActivationFunctions.Tanh, 10);
//network.InitializeWeights();
NeuralNetwork network = new NeuralNetwork(_dp, "Simard-6", 10, 0.8D, LossFunctions.MeanSquareError, DataProviderSets.MNIST, TrainingStrategy.SGDLevenbergMarquardt, 0.02D);
network.AddLayer(LayerTypes.Input, 1, 32, 32);
network.AddLayer(LayerTypes.ConvolutionalSubsampling, ActivationFunctions.Tanh, 6, 14, 14, 5, 5);
network.AddLayer(LayerTypes.ConvolutionalSubsampling, ActivationFunctions.Tanh, 50, 5, 5, 5, 5);
network.AddLayer(LayerTypes.FullyConnected, ActivationFunctions.Tanh, 100);
network.AddLayer(LayerTypes.FullyConnected, ActivationFunctions.Tanh, 10);
network.InitializeWeights();
//NeuralNetwork network = new NeuralNetwork(DataProvider, "Simard-16", 10, 0.8D, LossFunctions.MeanSquareError, DataProviderSets.MNIST, TrainingStrategy.SGDLevenbergMarquardt, 0.1D);
//network.AddLayer(LayerTypes.Input, 1, 32, 32);
//network.AddLayer(LayerTypes.ConvolutionalSubsampling, ActivationFunctions.Tanh, 16, 14, 14, 5, 5);
//network.AddLayer(LayerTypes.ConvolutionalSubsampling, ActivationFunctions.Tanh, 64, 5, 5, 5, 5);
//network.AddLayer(LayerTypes.FullyConnected, ActivationFunctions.Tanh, 196);
//network.AddLayer(LayerTypes.FullyConnected, ActivationFunctions.Tanh, 10);
//network.InitializeWeights();
//NeuralNetwork network = new NeuralNetwork(DataProvider, "LeNet-5", 10, 0.8D, LossFunctions.MeanSquareError, DataProviderSets.MNIST, TrainingStrategy.SGDLevenbergMarquardt, 0.02D);
//network.AddLayer(LayerTypes.Input, 1, 32, 32);
//network.AddLayer(LayerTypes.Convolutional, ActivationFunctions.Tanh, 6, 28, 28, 5, 5, 1, 1);
//network.AddLayer(LayerTypes.AvgPooling, ActivationFunctions.Tanh, 6, 14, 14, 2, 2, 2, 2);
//bool[] maps = new bool[6 * 16]
//{
// true, false, false, false, true, true, true, false, false, true, true, true, true, false, true, true,
// true, true, false, false, false, true, true, true, false, false, true, true, true, true, false, true,
// true, true, true, false, false, false, true, true, true, false, false, true, false, true, true, true,
// false, true, true, true, false, false, true, true, true, true, false, false, true, false, true, true,
// false, false, true, true, true, false, false, true, true, true, true, false, true, true, false, true,
// false, false, false, true, true, true, false, false, true, true, true, true, false, true, true, true
//};
//network.AddLayer(LayerTypes.Convolutional, ActivationFunctions.Tanh, 16, 10, 10, 5, 5, 1, 1, 0, 0, new Mappings(maps));
//network.AddLayer(LayerTypes.AvgPooling, ActivationFunctions.Tanh, 16, 5, 5, 2, 2, 2, 2);
//network.AddLayer(LayerTypes.Convolutional, ActivationFunctions.Tanh, 120, 1, 1, 5, 5, 1, 1);
//network.AddLayer(LayerTypes.FullyConnected, ActivationFunctions.Tanh, 10);
//network.InitializeWeights();
//NeuralNetwork network = new NeuralNetwork(DataProvider, "LeNet-5", 10, 1D, LossFunctions.CrossEntropy, DataProviderSets.MNIST, TrainingStrategy.SGDLevenbergMarquardt, 0.02D);
//network.AddLayer(LayerTypes.Input, 1, 32, 32);
//network.AddLayer(LayerTypes.Convolutional, ActivationFunctions.ReLU, 6, 28, 28, 5, 5);
//network.AddLayer(LayerTypes.AvgPooling, ActivationFunctions.Ident, 6, 14, 14, 2, 2, 2, 2);
//bool[] maps = new bool[6 * 16]
//{
// true, false, false, false, true, true, true, false, false, true, true, true, true, false, true, true,
// true, true, false, false, false, true, true, true, false, false, true, true, true, true, false, true,
// true, true, true, false, false, false, true, true, true, false, false, true, false, true, true, true,
// false, true, true, true, false, false, true, true, true, true, false, false, true, false, true, true,
// false, false, true, true, true, false, false, true, true, true, true, false, true, true, false, true,
// false, false, false, true, true, true, false, false, true, true, true, true, false, true, true, true
//};
//network.AddLayer(LayerTypes.Convolutional, ActivationFunctions.ReLU, 16, 10, 10, 5, 5, 1, 1, 0, 0, new Mappings(maps));
//network.AddLayer(LayerTypes.AvgPooling, ActivationFunctions.Ident, 16, 5, 5, 2, 2, 2, 2);
//network.AddLayer(LayerTypes.Local, ActivationFunctions.ReLU, 120, 1, 1, 5, 5);
//network.AddLayer(LayerTypes.FullyConnected, ActivationFunctions.SoftMax, 10);
//network.InitializeWeights();
//NeuralNetwork network = new NeuralNetwork(DataProvider, "MyNet-16", 10, 0.8D, LossFunctions.MeanSquareError, DataProviderSets.MNIST, TrainingStrategy.SGDLevenbergMarquardt, 0.02D);
//network.AddLayer(LayerTypes.Input, 1, 32, 32);
//network.AddLayer(LayerTypes.Convolutional, ActivationFunctions.Tanh, 16, 28, 28, 5, 5);
//network.AddLayer(LayerTypes.AvgPooling, ActivationFunctions.Tanh, 16, 14, 14, 2, 2);
//network.AddLayer(LayerTypes.Convolutional, ActivationFunctions.Tanh, 64, 10, 10, 5, 5, 1, 1, 0, 0, new Mappings(16, 64, 66, 1));
//network.AddLayer(LayerTypes.AvgPooling, ActivationFunctions.Tanh, 64, 5, 5, 2, 2);
//network.AddLayer(LayerTypes.Convolutional, ActivationFunctions.Tanh, 256, 1, 1, 5, 5, 1, 1, 0, 0, new Mappings(64, 256, 66, 2));
//network.AddLayer(LayerTypes.FullyConnected, ActivationFunctions.Tanh, 10);
//network.InitializeWeights();
//NeuralNetwork network = new NeuralNetwork(DataProvider, "CNN-CIFAR10-A", 10, 0.8D, LossFunctions.MeanSquareError, DataProviderSets.CIFAR10, TrainingStrategy.SGDLevenbergMarquardt, 0.02D);
//network.AddLayer(LayerTypes.Input, 3, 32, 32);
//bool[] maps = new bool[3 * 64]
//{
// true, true, true, true, true, true, true, true, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, true, true, true, true, true, true, true, true, true, true, true, true, true, true, true, true, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, true, true, true, true, true, true, true, true,
// false, false, false, false, false, false, false, false, true, true, true, true, true, true, true, true, false, false, false, false, false, false, false, false, true, true, true, true, true, true, true, true, false, false, false, false, false, false, false, false, true, true, true, true, true, true, true, true, false, false, false, false, false, false, false, false, true, true, true, true, true, true, true, true,
// false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, true, true, true, true, true, true, true, true, true, true, true, true, true, true, true, true,false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, true, true, true, true, true, true, true, true, true, true, true, true, true, true, true, true
//};
//network.AddLayer(LayerTypes.Convolutional, ActivationFunctions.ReLU, 64, 28, 28, 5, 5, 1, 1, 0, 0, new Mappings(maps));
//network.AddLayer(LayerTypes.StochasticPooling, ActivationFunctions.Ident, 64, 14, 14, 3, 3, 2, 2);
//network.AddLayer(LayerTypes.Convolutional, ActivationFunctions.ReLU, 64, 10, 10, 5, 5, 1, 1, 0, 0, new Mappings(64, 64, 66, 1));
//network.AddLayer(LayerTypes.StochasticPooling, ActivationFunctions.Ident, 64, 5, 5, 3, 3, 2, 2);
//network.AddLayer(LayerTypes.Local, ActivationFunctions.ReLU, 64, 3, 3, 3, 3, 1, 1, 0, 0, new Mappings(64, 64, 66, 2));
//network.AddLayer(LayerTypes.Local, ActivationFunctions.ReLU, 386, 1, 1, 3, 3, 1, 1, 0, 0, 50);
//network.AddLayer(LayerTypes.FullyConnected, ActivationFunctions.SoftSign, 10);
//network.InitializeWeights();
//NeuralNetwork network = new NeuralNetwork(DataProvider, "CNN-CIFAR10-Z2", 10, 1D, LossFunctions.CrossEntropy, DataProviderSets.CIFAR10, TrainingStrategy.SGDLevenbergMarquardt);
//network.AddLayer(LayerTypes.Input, 3, 32, 32);
//bool[] maps = new bool[3 * 64]
//{
// true, true, true, true, true, true, true, true, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, true, true, true, true, true, true, true, true, true, true, true, true, true, true, true, true, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, true, true, true, true, true, true, true, true,
// false, false, false, false, false, false, false, false, true, true, true, true, true, true, true, true, false, false, false, false, false, false, false, false, true, true, true, true, true, true, true, true, false, false, false, false, false, false, false, false, true, true, true, true, true, true, true, true, false, false, false, false, false, false, false, false, true, true, true, true, true, true, true, true,
// false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, true, true, true, true, true, true, true, true, true, true, true, true, true, true, true, true,false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, true, true, true, true, true, true, true, true, true, true, true, true, true, true, true, true
//};
//bool[] maps = new bool[3 * 96]
//{
// true, false, false, true, false, false, true, false, false, true, false, false, true, false, false, true, false, false, true, false, false, true, false, false, true, false, false, true, false, false, true, false, false, true, false, false, true, false, false, true, false, false, true, false, false, true, false, false, true, false, false, true, false, false, true, false, false, true, false, false, true, false, false, true, false, false, true, false, false, true, false, false, true, false, false, true, false, false, true, false, false, true, false, false, true, false, false, true, false, false, true, false, false, true, false, false,
// false, true, false, false, true, false, false, true, false, false, true, false, false, true, false, false, true, false, false, true, false, false, true, false, false, true, false, false, true, false, false, true, false, false, true, false, false, true, false, false, true, false, false, true, false, false, true, false, false, true, false, false, true, false, false, true, false, false, true, false, false, true, false, false, true, false, false, true, false, false, true, false, false, true, false, false, true, false, false, true, false, false, true, false, false, true, false, false, true, false, false, true, false, false, true, false,
// false, false, true, false, false, true, false, false, true, false, false, true, false, false, true, false, false, true, false, false, true, false, false, true, false, false, true, false, false, true, false, false, true, false, false, true, false, false, true, false, false, true, false, false, true, false, false, true, false, false, true, false, false, true, false, false, true, false, false, true, false, false, true, false, false, true, false, false, true, false, false, true, false, false, true, false, false, true, false, false, true, false, false, true, false, false, true, false, false, true, false, false, true, false, false, true
//};
//network.AddLayer(LayerTypes.Convolutional, ActivationFunctions.ReLU, 64, 28, 28, 5, 5, 1, 1, 0, 0, new Mappings(maps));
//network.AddLayer(LayerTypes.StochasticPooling, ActivationFunctions.Ident, 64, 14, 14, 3, 3, 2, 2);
////network.AddLayer(LayerTypes.LocalResponseNormalizationCM, ActivationFunctions.None, 64, 14, 14, 3, 3);
//network.AddLayer(LayerTypes.Convolutional, ActivationFunctions.ReLU, 64, 10, 10, 5, 5, 1, 1, 0, 0, new Mappings(64, 64, 66, 1));
////network.AddLayer(LayerTypes.LocalResponseNormalizationCM, ActivationFunctions.None, 64, 10, 10, 3, 3);
//network.AddLayer(LayerTypes.StochasticPooling, ActivationFunctions.Ident, 64, 5, 5, 3, 3, 2, 2);
////network.AddLayer(LayerTypes.Local, ActivationFunctions.ReLU, 64, 1, 1, 5, 5, 1, 1, 0, 0, new Mappings(64, 64, 66, 2));
//network.AddLayer(LayerTypes.Local, ActivationFunctions.Logistic, 384, 1, 1, 5, 5, 1, 1, 0, 0, 50);
//network.AddLayer(LayerTypes.FullyConnected, ActivationFunctions.SoftMax, 10);
//network.InitializeWeights();
//network.LoadWeights(StorageDirectory + @"\CNN-CIFAR10-Z2 (2259 errors).weights-bin");
//NeuralNetwork network = new NeuralNetwork(DataProvider, "CNN-CIFAR10-B2", 10, 0.8D, LossFunctions.MeanSquareError, DataProviderSets.CIFAR10, TrainingStrategy.SGDLevenbergMarquardt, 0.02D);
//network.AddLayer(LayerTypes.Input, 3, 32, 32);
//bool[] maps = new bool[3 * 64]
//{
// true, true, true, true, true, true, true, true, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, true, true, true, true, true, true, true, true, true, true, true, true, true, true, true, true, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, true, true, true, true, true, true, true, true,
// false, false, false, false, false, false, false, false, true, true, true, true, true, true, true, true, false, false, false, false, false, false, false, false, true, true, true, true, true, true, true, true, false, false, false, false, false, false, false, false, true, true, true, true, true, true, true, true, false, false, false, false, false, false, false, false, true, true, true, true, true, true, true, true,
// false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, true, true, true, true, true, true, true, true, true, true, true, true, true, true, true, true,false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, true, true, true, true, true, true, true, true, true, true, true, true, true, true, true, true
//};
//network.AddLayer(LayerTypes.Convolutional, ActivationFunctions.ReLU, 64, 28, 28, 5, 5, 1, 1, 0, 0, new Mappings(maps));
//network.AddLayer(LayerTypes.StochasticPooling, ActivationFunctions.Ident, 64, 14, 14, 3, 3, 2, 2);
//network.AddLayer(LayerTypes.Convolutional, ActivationFunctions.ReLU, 64, 10, 10, 5, 5, 1, 1, 0, 0, new Mappings(64, 64, 66, 1));
//network.AddLayer(LayerTypes.StochasticPooling, ActivationFunctions.Ident, 64, 5, 5, 3, 3, 2, 2);
//network.AddLayer(LayerTypes.Local, ActivationFunctions.Tanh, 512, 1, 1, 5, 5, 1, 1, 0, 0, 50);
//network.AddLayer(LayerTypes.FullyConnected, ActivationFunctions.Tanh, 10);
//network.InitializeWeights();
//NeuralNetwork network = new NeuralNetwork(DataProvider, "CNN-CIFAR10-Z3", 10, 1D, LossFunctions.CrossEntropy, DataProviderSets.CIFAR10, TrainingStrategy.SGDLevenbergMarquardt, 0.02D);
//network.AddLayer(LayerTypes.Input, 3, 32, 32);
//bool[] maps = new bool[3 * 64]
//{
// true, true, true, true, true, true, true, true, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, true, true, true, true, true, true, true, true, true, true, true, true, true, true, true, true, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, true, true, true, true, true, true, true, true,
// false, false, false, false, false, false, false, false, true, true, true, true, true, true, true, true, false, false, false, false, false, false, false, false, true, true, true, true, true, true, true, true, false, false, false, false, false, false, false, false, true, true, true, true, true, true, true, true, false, false, false, false, false, false, false, false, true, true, true, true, true, true, true, true,
// false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, true, true, true, true, true, true, true, true, true, true, true, true, true, true, true, true,false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, true, true, true, true, true, true, true, true, true, true, true, true, true, true, true, true
//};
//network.AddLayer(LayerTypes.Convolutional, ActivationFunctions.ReLU, 64, 32, 32, 5, 5, 1, 1, 2, 2, new Mappings(maps));
//network.AddLayer(LayerTypes.StochasticPooling, ActivationFunctions.Ident, 64, 16, 16, 3, 3, 2, 2);
////network.AddLayer(LayerTypes.LocalResponseNormalizationCM, ActivationFunctions.None, 64, 16, 16, 5, 5);
//network.AddLayer(LayerTypes.Convolutional, ActivationFunctions.ReLU, 64, 16, 16, 5, 5, 1, 1, 2, 2, new Mappings(64, 64, 66, 1));
////network.AddLayer(LayerTypes.LocalResponseNormalizationCM, ActivationFunctions.None, 64, 16, 16, 5, 5);
//network.AddLayer(LayerTypes.StochasticPooling, ActivationFunctions.Ident, 64, 8, 8, 3, 3, 2, 2);
//network.AddLayer(LayerTypes.Local, ActivationFunctions.ReLU, 32, 8, 8, 3, 3, 1, 1, 1, 1, new Mappings(64, 32, 50, 2));
//network.AddLayer(LayerTypes.Local, ActivationFunctions.SoftSign, 32, 8, 8, 3, 3, 1, 1, 1, 1, 50);
//network.AddLayer(LayerTypes.FullyConnected, ActivationFunctions.SoftMax, 10);
//network.InitializeWeights();
//NeuralNetwork network = new NeuralNetwork(DataProvider, "CNN-CIFAR10-C", 10, 0.8D, LossFunctions.MeanSquareError, DataProviderSets.CIFAR10, TrainingStrategy.SGDLevenbergMarquardt, 0.02D);
//network.AddLayer(LayerTypes.Input, 3, 32, 32);
//bool[] maps = new bool[3 * 64]
//{
// true, true, true, true, true, true, true, true, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, true, true, true, true, true, true, true, true, true, true, true, true, true, true, true, true, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, true, true, true, true, true, true, true, true,
// false, false, false, false, false, false, false, false, true, true, true, true, true, true, true, true, false, false, false, false, false, false, false, false, true, true, true, true, true, true, true, true, false, false, false, false, false, false, false, false, true, true, true, true, true, true, true, true, false, false, false, false, false, false, false, false, true, true, true, true, true, true, true, true,
// false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, true, true, true, true, true, true, true, true, true, true, true, true, true, true, true, true,false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, true, true, true, true, true, true, true, true, true, true, true, true, true, true, true, true
//};
//network.AddLayer(LayerTypes.Convolutional, ActivationFunctions.ReLU, 64, 28, 28, 5, 5, 1, 1, 0, 0, new Mappings(maps));
//network.AddLayer(LayerTypes.StochasticPooling, ActivationFunctions.Ident, 64, 14, 14, 3, 3, 2, 2);
//network.AddLayer(LayerTypes.Convolutional, ActivationFunctions.ReLU, 64, 10, 10, 5, 5, 1, 1, 0, 0, new Mappings(64, 64, 66, 1));
//network.AddLayer(LayerTypes.StochasticPooling, ActivationFunctions.Ident, 64, 5, 5, 3, 3, 2, 2);
//network.AddLayer(LayerTypes.Local, ActivationFunctions.ReLU, 32, 5, 5, 3, 3, 1, 1, 1, 1, new Mappings(64, 32, 50, 2));
//network.AddLayer(LayerTypes.Local, ActivationFunctions.ReLU, 32, 5, 5, 3, 3, 1, 1, 1, 1, 50);
//network.AddLayer(LayerTypes.FullyConnected, ActivationFunctions.Tanh, 10);
//network.InitializeWeights();
//NeuralNetwork network = new NeuralNetwork(DataProvider, "CNN-CIFAR10-Z4", 10, 0.8D, LossFunctions.MeanSquareError, DataProviderSets.CIFAR10, TrainingStrategy.SGDLevenbergMarquardt, 0.02D, 2000, false);
//network.AddLayer(LayerTypes.Input, 3, 32, 32);
////bool[] maps = new bool[3 * 64]
////{
//// true, true, true, true, true, true, true, true, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, true, true, true, true, true, true, true, true, true, true, true, true, true, true, true, true, false, false, false, false, false, false, false, false, true, true, true, true, true, true, true, true, false, true, false, true, false, true, false, true,
//// false, false, false, false, false, false, false, false, true, true, true, true, true, true, true, true, false, false, false, false, false, false, false, false, true, true, true, true, true, true, true, true, true, true, true, true, true, true, true, true, true, true, true, true, true, true, true, true, true, true, true, true, true, true, true, true, true, false, true, false, true, false, true, false,
//// false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, true, true, true, true, true, true, true, true, true, true, true, true, true, true, true, true, false, false, false, false, false, false, false, false, true, true, true, true, true, true, true, true, true, true, true, true, true, true, true, true, false, true, false, true, false, true, false, true
////};
//bool[] maps = new bool[3 * 64]
//{
// true, true, true, true, true, true, true, true, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, true, true, true, true, true, true, true, true, true, true, true, true, true, true, true, true, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, true, true, true, true, true, true, true, true,
// false, false, false, false, false, false, false, false, true, true, true, true, true, true, true, true, false, false, false, false, false, false, false, false, true, true, true, true, true, true, true, true, false, false, false, false, false, false, false, false, true, true, true, true, true, true, true, true, false, false, false, false, false, false, false, false, true, true, true, true, true, true, true, true,
// false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, true, true, true, true, true, true, true, true, true, true, true, true, true, true, true, true,false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, true, true, true, true, true, true, true, true, true, true, true, true, true, true, true, true
//};
////bool[] maps = new bool[3 * 48] //Z3
////{
//// true, true, true, true, true, true, true, true, true, true, true, true, true, true, true, true, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false,
//// false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, true, true, true, true, true, true, true, true, true, true, true, true, true, true, true, true, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false,
//// false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, true, true, true, true, true, true, true, true, true, true, true, true, true, true, true, true
////};
////bool[] maps = new bool[3 * 48] //Z3
////{
//// true, true, true, true, true, true, true, true, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, true, true, true, true, true, true, true, true, true, true, true, true, true, true, true, true, false, false, false, false, false, false, false, false,
//// false, false, false, false, false, false, false, false, true, true, true, true, true, true, true, true, false, false, false, false, false, false, false, false, true, true, true, true, true, true, true, true, true, true, true, true, true, true, true, true, true, true, true, true, true, true, true, true,
//// false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, true, true, true, true, true, true, true, true, true, true, true, true, true, true, true, true, false, false, false, false, false, false, false, false, true, true, true, true, true, true, true, true
////};
//network.AddLayer(LayerTypes.Convolutional, ActivationFunctions.ReLU, 64, 32, 32, 5, 5, 1, 1, 2, 2, new Mappings(maps));
//network.AddLayer(LayerTypes.StochasticPooling, ActivationFunctions.Ident, 64, 16, 16, 3, 3, 2, 2);
////network.AddLayer(LayerTypes.LocalContrastNormalization, ActivationFunctions.None, 64, 16, 16, 5, 5);
//network.AddLayer(LayerTypes.Convolutional, ActivationFunctions.ReLU, 64, 16, 16, 5, 5, 1, 1, 2, 2, new Mappings(64, 64, 66, 1));
////network.AddLayer(LayerTypes.LocalContrastNormalization, ActivationFunctions.None, 64, 16, 16, 5, 5);
//network.AddLayer(LayerTypes.StochasticPooling, ActivationFunctions.Ident, 64, 8, 8, 3, 3, 2, 2);
//network.AddLayer(LayerTypes.Local, ActivationFunctions.ReLU, 32, 8, 8, 3, 3, 1, 1, 1, 1, new Mappings(64, 32, 50, 2));
//network.AddLayer(LayerTypes.Local, ActivationFunctions.ReLU, 32, 8, 8, 3, 3, 1, 1, 1, 1, 50);
//network.AddLayer(LayerTypes.FullyConnected, ActivationFunctions.SoftSign, 10);
//network.InitializeWeights();
//NeuralNetwork network = new NeuralNetwork(DataProvider, "CNN-CIFAR10-D", 10, 0.8D, LossFunctions.MeanSquareError, DataProviderSets.CIFAR10, TrainingStrategy.SGDLevenbergMarquardt, 0.02D);
//network.AddLayer(LayerTypes.Input, 3, 32, 32);
//bool[] maps = new bool[3 * 64]
//{
// true, true, true, true, true, true, true, true, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, true, true, true, true, true, true, true, true, true, true, true, true, true, true, true, true, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, true, true, true, true, true, true, true, true,
// false, false, false, false, false, false, false, false, true, true, true, true, true, true, true, true, false, false, false, false, false, false, false, false, true, true, true, true, true, true, true, true, false, false, false, false, false, false, false, false, true, true, true, true, true, true, true, true, false, false, false, false, false, false, false, false, true, true, true, true, true, true, true, true,
// false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, true, true, true, true, true, true, true, true, true, true, true, true, true, true, true, true,false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, true, true, true, true, true, true, true, true, true, true, true, true, true, true, true, true
//};
//network.AddLayer(LayerTypes.Convolutional, ActivationFunctions.ReLU, 64, 32, 32, 5, 5, 1, 1, 2, 2, new Mappings(maps));
//network.AddLayer(LayerTypes.MaxPoolingWeightless, ActivationFunctions.Ident, 64, 16, 16, 3, 3, 2, 2);
//network.AddLayer(LayerTypes.Convolutional, ActivationFunctions.ReLU, 64, 16, 16, 3, 3, 1, 1, 1, 1, new Mappings(64, 64, 50, 1));
//network.AddLayer(LayerTypes.AvgPoolingWeightless, ActivationFunctions.Ident, 64, 8, 8, 3, 3, 2, 2);
//network.AddLayer(LayerTypes.Local, ActivationFunctions.BReLU, 32, 8, 8, 3, 3, 1, 1, 1, 1, new Mappings(64, 32, 50, 2));
//network.AddLayer(LayerTypes.Local, ActivationFunctions.Tanh, 32, 8, 8, 3, 3, 1, 1, 1, 1, 50);
//network.AddLayer(LayerTypes.FullyConnected, ActivationFunctions.Tanh, 10);
////network.InitializeWeights();
//NeuralNetwork network = new NeuralNetwork(DataProvider, "CNN-CIFAR10-G", 10, 0.8D, LossFunctions.MeanSquareError, DataProviderSets.CIFAR10, TrainingStrategy.SGD, 0.02D);
//network.AddLayer(LayerTypes.Input, 3, 32, 32);
//bool[] maps = new bool[3 * 48]
//{
// true, true, true, true, true, true, true, true, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, true, true, true, true, true, true, true, true, true, true, true, true, true, true, true, true, false, false, false, false, false, false, false, false,
// false, false, false, false, false, false, false, false, true, true, true, true, true, true, true, true, false, false, false, false, false, false, false, false, true, true, true, true, true, true, true, true, true, true, true, true, true, true, true, true, true, true, true, true, true, true, true, true,
// false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, true, true, true, true, true, true, true, true, true, true, true, true, true, true, true, true, false, false, false, false, false, false, false, false, true, true, true, true, true, true, true, true
//};
//network.AddLayer(LayerTypes.Convolutional, ActivationFunctions.ReLU, 48, 32, 32, 5, 5, 1, 1, 2, 2, new Mappings(maps));
//network.AddLayer(LayerTypes.StochasticPooling, ActivationFunctions.Ident, 48, 16, 16, 3, 3, 2, 2);
//network.AddLayer(LayerTypes.Convolutional, ActivationFunctions.ReLU, 64, 16, 16, 5, 5, 1, 1, 2, 2, new Mappings(48, 64, 66, 1));
//network.AddLayer(LayerTypes.StochasticPooling, ActivationFunctions.Ident, 64, 8, 8, 3, 3, 2, 2);
//network.AddLayer(LayerTypes.Local, ActivationFunctions.ReLU, 24, 8, 8, 3, 3, 1, 1, 1, 1, new Mappings(64, 24, 50, 2));
//network.AddLayer(LayerTypes.Local, ActivationFunctions.Tanh, 24, 8, 8, 3, 3, 1, 1, 1, 1, 50);
//network.AddLayer(LayerTypes.FullyConnected, ActivationFunctions.Tanh, 10);
//network.InitializeWeights();
//NeuralNetwork network = new NeuralNetwork(DataProvider, "CNN-MNIST-A", 10, 1D, LossFunctions.CrossEntropy, DataProviderSets.MNIST, TrainingStrategy.SGD, 0.02D);
//network.AddLayer(LayerTypes.Input, 1, 32, 32);
//network.AddLayer(LayerTypes.Convolutional, ActivationFunctions.ReLU, 32, 24, 24, 9, 9, 1, 1, 0, 0);
//network.AddLayer(LayerTypes.StochasticPooling, ActivationFunctions.Ident, 32, 12, 12, 3, 3, 2, 2);
//network.AddLayer(LayerTypes.Convolutional, ActivationFunctions.ReLU, 32, 8, 8, 5, 5, 1, 1, 0, 0, new Mappings(32, 32, 66));
//network.AddLayer(LayerTypes.StochasticPooling, ActivationFunctions.Ident, 32, 4, 4, 3, 3, 2, 2);
//network.AddLayer(LayerTypes.Local, ActivationFunctions.ReLU, 256, 1, 1, 4, 4, 1, 1, 0, 0, 50);
//network.AddLayer(LayerTypes.FullyConnected, ActivationFunctions.SoftMax, 10);
//network.InitializeWeights();
//network.RaiseNetworkProgressEvent += new EventHandler<EventArgs>(NetworkProgressEvent);
//network.RaiseAddUnrecognizedTestSampleEvent += new EventHandler<AddUnrecognizedTestSampleEventArgs>(AddUnrecognizedTestSampleEvent);
network.MaxDegreeOfParallelism = 8;
network.AddGlobalTrainingRate(_data, true);
return(network);
}
public static void Main()
{
int height = 100;
int width = 100;
float[] input = new float[height * width];
for (int i = 0; i < input.Length; i++)
{
input[i] = .5f;
}
Random rnd = new Random();
InOutPair[] IO = new InOutPair[1];
float[] toAdd = new float[height * width];
for (int i = 0; i < height * width; i++)
{
toAdd[i] = 1f / i;
}
IO[0] = new InOutPair(toAdd, new float[] { .1f, .5f, -.1f });
Epoch Ep = new Epoch(IO, 1000, true);
NeuralNetwork NN = new NeuralNetwork(9, rnd, .01f, 1, 10, 200);
NN.AddLayer(5, "HypTan");
NN.AddLayer(5, "HypTan");
NN.AddLayer(5, "HypTan");
NN.AddLayer(3, "HypTan");
Convolution CC = new Convolution(height, width);
CC.AddLayer(5, 5, 5);
CC.AddLayer(2, 2, 1, new HyperbolicTangent());
CC.AddLayer(6, 6, 1);
CC.AddLayer(2, 2, 2, true);
CC.AddLayer(2, 2, 1, new HyperbolicTangent());
CC.AddLayer(2, 2, 2, true);
Combination C = new Combination(CC, NN);
float[] layerOut = C.Run(input);
for (int i = 0; i < layerOut.Length; i++)
{
Console.WriteLine(layerOut[i]);
}
C.Train(Ep);
Console.WriteLine();
layerOut = C.Run(input);
for (int i = 0; i < layerOut.Length; i++)
{
Console.WriteLine(layerOut[i]);
}
}
static void Main(string[] args)
{
Console.WriteLine("Hello World!");
NeuralNetwork net = new NeuralNetwork();
NeuralLayer LI = new NeuralLayer("LI");
//LI.AddNeuron(new StepNeuron(1, 1));
//LI.AddNeuron(new StepNeuron(1, 1));
LI.AddNeuron(new InputNeuron());
LI.AddNeuron(new InputNeuron());
net.AddLayer(LI);
NeuralLayer H0 = new NeuralLayer("H0");
H0.AddNeuron(new StepNeuron(2, 0.1));
H0.AddNeuron(new StepNeuron(2, 0.1));
net.AddLayer(H0);
NeuralLayer H1 = new NeuralLayer("H1");
H1.AddNeuron(new StepNeuron(2, 0.1));
H1.AddNeuron(new StepNeuron(2, 0.1));
net.AddLayer(H1);
//NeuralLayer H2 = new NeuralLayer("H1");
//H2.AddNeuron(new StepNeuron(2, 0.2));
//H2.AddNeuron(new StepNeuron(2, 0.2));
//net.AddLayer(H2);
NeuralLayer LO = new NeuralLayer("LO");
LO.AddNeuron(new StepNeuron(2, 0.1));
net.AddLayer(LO);
//NeuralLayer LI = new NeuralLayer("LI");
//LI.AddNeuron(new StepNeuron(2, 0.1));
//net.AddLayer(LI);
//for (int i = 0; i < 10; i++)
//{
// net.SetInputs(0, 0);
// net.Compute();
// Console.WriteLine($"output: {net.GetOutputs().First().Value} error: {net.CalculateTotalError(0)}");
// net.SetInputs(0, 1);
// net.Compute();
// Console.WriteLine($"output: {net.GetOutputs().First().Value} error: {net.CalculateTotalError(0)}");
// net.SetInputs(1, 0);
// net.Compute();
// Console.WriteLine($"output: {net.GetOutputs().First().Value} error: {net.CalculateTotalError(0)}");
// net.SetInputs(1, 1);
// net.Compute();
// Console.WriteLine($"output: {net.GetOutputs().First().Value} error: {net.CalculateTotalError(1)}");
// Console.WriteLine("-------------");
// net.AdjustWeight(.1, 1);
//}
var inputs = new List <List <double> >()
{
new List <double> {
0, 0
},
new List <double> {
0, 1
},
new List <double> {
1, 0
},
new List <double> {
1, 1
},
};
var expectedValues = new List <List <double> >()
{
new List <double> {
0
},
new List <double> {
0
},
new List <double> {
0
},
new List <double> {
1
},
};
net.Train(inputs, expectedValues, 0.1, 10);
foreach (var layer in net.Layers)
{
Console.Write($"{layer.LayerName} {layer.Neurons.Count}");
foreach (var neuron in layer.Neurons)
{
foreach (var dendrite in neuron.Dendrites)
{
Console.Write($", {dendrite.Weight}");
}
}
Console.WriteLine();
}
}
public override void OnInspectorGUI()
{
#region Gizmo Options
if (GUILayout.Button(showingGizmoOptions ? "Hide Gizmo Options" : "Show Gizmo Options"))
{
showingGizmoOptions = !showingGizmoOptions;
if (showingGizmoOptions)
{
nodeSize = NeuralNetwork.Editor.nodeSize;
nSpace = nodeSpacing;
lSpace = layerSpacing;
color = nodeColor;
gradient = axonGradient;
}
}
if (showingGizmoOptions)
{
EditorGUI.indentLevel = 1;
color = EditorGUILayout.ColorField("Node Color", color);
gradient = EditorGUILayout.GradientField("Relation Gradient", gradient);
nodeSize = EditorGUILayout.FloatField("Node Size", nodeSize);
nSpace = EditorGUILayout.FloatField("Node Spacing", nSpace);
lSpace = EditorGUILayout.FloatField("Layer Spacing", lSpace);
if (GUILayout.Button("Reset"))
{
ApplyGizmoSettings(Color.cyan, GetDefaultGradient(), 1.0f, 5.0f, 5.0f);
}
if (GUILayout.Button("Apply"))
{
ApplyGizmoSettings(color, gradient, nodeSize, nSpace, lSpace);
}
EditorGUI.indentLevel = 0;
}
#endregion
GUILayout.Space(15);
#region I/O Nodes
int prevVal = 0;
bool nodeAdded = false;
var editorProp = serializedObject.FindProperty("editorUtil");
var layersProp = editorProp.FindPropertyRelative("nodeHeights");
modifiedLayers.Clear();
nodeAdded |= NodeLayerMod(layersProp.GetArrayElementAtIndex(0), 0, "Num Input Nodes");
nodeAdded |= NodeLayerMod(layersProp.GetArrayElementAtIndex(layersProp.arraySize - 1), layersProp.arraySize - 1, "Num Output Nodes");
#endregion
GUILayout.Space(5);
#region Hidden Layers
int layerRem = -1;
bool layerAdd = false;
EditorGUI.indentLevel = 1;
if (GUILayout.Button("Add Hidden Layer"))
{
layerAdd = true;
int v = layersProp.GetArrayElementAtIndex(layersProp.arraySize - 2).intValue;
layersProp.InsertArrayElementAtIndex(layersProp.arraySize - 1);
layersProp.GetArrayElementAtIndex(layersProp.arraySize - 2).intValue = v;
}
for (int i = 1; i < layersProp.arraySize - 1; ++i)
{
var indexProp = layersProp.GetArrayElementAtIndex(i);
nodeAdded |= NodeLayerMod(indexProp, i, $"Hidden Layer {i.ToString()} Nodes", out bool remove);
if (remove)
{
layerRem = i;
layersProp.DeleteArrayElementAtIndex(i--);
}
}
EditorGUI.indentLevel = 0;
#endregion
GUILayout.Space(15);
serializedObject.ApplyModifiedProperties();
if (nodeAdded)
{
Debug.Log("Refreshing gizmo data");
foreach (var v2i in modifiedLayers)
{
if (v2i.y < 0)
{
nn.RemoveNodeFromLayer(v2i.x);
}
else
{
nn.AddNodeToLayer(v2i.x, false);
}
}
CacheGizmoDrawData(nn);
EditorUtility.SetDirty(nn);
}
if (layerAdd)
{
var p = layersProp.GetArrayElementAtIndex(layersProp.arraySize - 2);
nn.AddLayer(layersProp.arraySize - 2, p.intValue, false, false);
CacheGizmoDrawData(nn);
nn.RefreshConnections();
EditorUtility.SetDirty(nn);
}
if (layerRem > -1)
{
nn.RemoveLayer(layerRem, false);
CacheGizmoDrawData(nn);
nn.RefreshConnections();
EditorUtility.SetDirty(nn);
}
if (GUILayout.Button("Rebuild Network"))
{
ApplyNetworkChanges();
EditorUtility.SetDirty(nn);
}
if (GUILayout.Button("Reroll Connections"))
{
nn.ConnectNetworkFresh();
EditorUtility.SetDirty(nn);
}
GUILayout.Space(5);
if (GUILayout.Button("Refresh Connections"))
{
nn.RefreshConnections();
EditorUtility.SetDirty(nn);
}
}
public void Run()
{
Stopwatch watch = Stopwatch.StartNew();
NeuralNetwork network = new NeuralNetwork();
network.AddLayer("input", new InputLayer(2), BaseLayer.TYPE.INPUT);
network.AddLayer("hidden", new CoreLayer(8, ACTIVATION.SIGMOID, BaseLayer.TYPE.HIDDEN), BaseLayer.TYPE.HIDDEN);
network.AddLayer("output", new CoreLayer(1, ACTIVATION.SIGMOID, BaseLayer.TYPE.OUTPUT), BaseLayer.TYPE.OUTPUT);
network.AddConnection("input", "hidden", Connection.INIT.GLOROT_UNIFORM);
network.AddConnection("hidden", "output", Connection.INIT.GLOROT_UNIFORM);
/*
* Optimizer optimizer = new BackProp(network, 1e-5f, 0.99f, true)
* {
* Alpha = 0.1f
* };
*/
Optimizer optimizer = new RMSProp(network)
{
Alpha = 0.1f
};
optimizer.InitBatchMode(4);
Vector[] input = new Vector[4];
Vector[] target = new Vector[4];
//Vector output = null;
input[0] = Vector.Build(2, new float[] { 0f, 0f });
input[1] = Vector.Build(2, new float[] { 0f, 1f });
input[2] = Vector.Build(2, new float[] { 1f, 0f });
input[3] = Vector.Build(2, new float[] { 1f, 1f });
target[0] = Vector.Build(1, new float[] { 0f });
target[1] = Vector.Build(1, new float[] { 1f });
target[2] = Vector.Build(1, new float[] { 1f });
target[3] = Vector.Build(1, new float[] { 0f });
for (int e = 0; e < 200; e++)
{
//Console.Write("Start ");
//BasePool.Instance.Check();
float err = 0;
for (int i = 0; i < 4; i++)
{
err += optimizer.Train(input[i], target[i]);
}
Console.WriteLine(err);
//Console.Write("End ");
//BasePool.Instance.Check();
}
Console.WriteLine();
for (int i = 0; i < 4; i++)
{
Console.WriteLine(network.Activate(input[i])[0]);
Vector.Release(input[i]);
Vector.Release(target[i]);
}
optimizer.Dispose();
Console.Write("Finish ");
BasePool.Instance.Check();
watch.Stop();
Console.WriteLine(watch.ElapsedMilliseconds);
}