public void Initialise()
{
layers[0] = LayerBuilder.CreateInputLayer(layers[0], layers[1]);
layers = LayerBuilder.CreateHidenLayer(layers);
layers[layers.Length - 1] = LayerBuilder.CreateOutputLayer(layers[layers.Length - 1],
layers[layers.Length - 2]);
}
// Constructs a convolutional network with two convolutional layers,
// two fully-connected layers, ReLU units and a softmax on the output.
public static void BuildCnn(IAllocator allocator, SeedSource seedSource, int batchSize, bool useCudnn, out Sequential model, out ICriterion criterion, out bool outputIsClassIndices)
{
var inputWidth = MnistParser.ImageSize;
var inputHeight = MnistParser.ImageSize;
var elementType = DType.Float32;
var inputDims = new long[] { batchSize, 1, inputHeight, inputWidth };
model = new Sequential();
model.Add(new ViewLayer(inputDims));
var outSize = AddCnnLayer(allocator, seedSource, elementType, model, inputDims, 20, useCudnn);
outSize = AddCnnLayer(allocator, seedSource, elementType, model, outSize, 40, useCudnn);
var convOutSize = outSize[1] * outSize[2] * outSize[3];
model.Add(new ViewLayer(batchSize, convOutSize));
var hiddenSize = 1000;
var outputSize = 10;
model.Add(new DropoutLayer(allocator, seedSource, elementType, 0.5f, batchSize, convOutSize));
model.Add(new LinearLayer(allocator, seedSource, elementType, (int)convOutSize, hiddenSize, batchSize));
model.Add(new ReLULayer(allocator, elementType, batchSize, hiddenSize));
model.Add(new DropoutLayer(allocator, seedSource, elementType, 0.5f, batchSize, hiddenSize));
model.Add(new LinearLayer(allocator, seedSource, elementType, hiddenSize, outputSize, batchSize));
model.Add(LayerBuilder.BuildLogSoftMax(allocator, elementType, batchSize, outputSize, useCudnn));
criterion = new ClassNLLCriterion(allocator, batchSize, outputSize);
outputIsClassIndices = true; // output of criterion is class indices
}
internal static ServerModelNode AddServerToHomeView(DappleModel oModel, LayerBuilder oLayer)
{
const bool Enabled = true;
const bool DontAddToHomeViewYet = false;
const bool DontSubmitToDappleSearch = false;
ServerModelNode result = null;
// --- Add the server to the model ---
if (oLayer is ArcIMSQuadLayerBuilder)
{
ArcIMSQuadLayerBuilder castLayer = oLayer as ArcIMSQuadLayerBuilder;
result = oModel.AddArcIMSServer(new ArcIMSServerUri(castLayer.ServerURL), Enabled, DontAddToHomeViewYet, DontSubmitToDappleSearch);
}
else if (oLayer is DAPQuadLayerBuilder)
{
DAPQuadLayerBuilder castLayer = oLayer as DAPQuadLayerBuilder;
result = oModel.AddDAPServer(new DapServerUri(castLayer.ServerURL), Enabled, DontAddToHomeViewYet, DontSubmitToDappleSearch);
}
else if (oLayer is WMSQuadLayerBuilder)
{
WMSQuadLayerBuilder castLayer = oLayer as WMSQuadLayerBuilder;
result = oModel.AddWMSServer(new WMSServerUri(castLayer.ServerURL), Enabled, DontAddToHomeViewYet, DontSubmitToDappleSearch);
}
else
{
throw new ApplicationException("Don't know how to get the server of type " + oLayer.GetType().ToString());
}
result.AddToHomeView();
return(result);
}
public void LayerAssembler_AssembleSimpleLayerWithRoomOffset_LayerCorrect()
{
var instructions = LayerInstruction.FromStrings(CreateLines("Layer:Size=(5,5)", "Room:(2,2)", "Floor:(1,1)"));
var layer = LayerBuilder.Assemble(instructions);
Assert.AreEqual("Floor", layer[3, 3].Ground.Type);
}
public void Init()
{
_routeProposal = new PatrolRouteProposal(_map, new XYZ(0, 0, 0), path => { });
var builder = new LayerBuilder(3, 3);
builder.PutFloor(new XY(0, 0), new XY(2, 2));
_map.Add(_layer = builder.Build());
}
public void LayerBuilder_CreateEmptyLayer_LayerCreated()
{
var builder = new LayerBuilder(10, 10);
var layer = builder.Build();
Assert.AreEqual(new XY(10, 10), layer.Size);
}
public void LayerBuilder_PutSingleFloorSpace_FloorCorrectlyAdded()
{
var builder = new LayerBuilder(3, 3);
builder.PutFloor(1, 1);
var layer = builder.Build();
AssertContains(1, 1, layer, FacilityObjectNames.Floor);
}
public void LayerBuilder_PutFloor_FloorWithCorrectObjectLayer()
{
var builder = new LayerBuilder(3, 3);
builder.PutFloor(1, 1);
var layer = builder.Build();
Assert.IsTrue(layer[1, 1].Ground.ObjectLayer == ObjectLayer.Ground);
}
public void LayerBuilder_PutFloorRectangle_FloorCorrectlyAdded()
{
var builder = new LayerBuilder(3, 3);
builder.PutFloor(new XY(0, 0), new XY(2, 2));
var layer = builder.Build();
AssertContains(new XY(0, 0), new XY(2, 2), layer, FacilityObjectNames.Floor);
}
public void LayerAssembler_AssembleSimpleLayerWithObjectLink_LayerCorrect()
{
var instructions = LayerInstruction.FromStrings(CreateLines("Layer:Size=(3,3)",
"Room: (0,0)", "Table:(1,1)", "Table:(1,2)", "Link:(Table,1,1)-(Table,1,2)"));
var layer = LayerBuilder.Assemble(instructions);
Assert.AreEqual(1, layer[1, 1].LowerObject.LinkedObjs.Count);
Assert.AreEqual(1, layer[1, 2].LowerObject.LinkedObjs.Count);
}
public void Init()
{
_thief = new Thief(this, _map);
_thief.Subscribe(this);
var builder = new LayerBuilder(3, 3);
builder.PutFloor(new XY(0, 0), new XY(2, 2));
AddPortals(builder);
_map.Add(_layer = builder.Build());
_upFacingContainer.Put(_upFacingValuable);
_rightFacingContainer.Put(_valuable2);
}
public void LayerAssembler_PutObjectFromInstruction_LayerCorrect()
{
var inst = ObjectInstruction.FromString("Floor:(1,1,R)");
var builder = new LayerBuilder(3, 3);
builder.Put(inst[0]);
var layer = builder.Build();
Assert.AreEqual("Floor", layer[1, 1].Ground.Type);
Assert.AreEqual(Orientation.Right, layer[1, 1].Ground.Orientation);
}
public void LayerBuilder_OnBuild_WallsCorrectlyAdded()
{
var builder = new LayerBuilder(3, 3);
builder.PutFloor(1, 1);
var layer = builder.Build();
AssertContains(new XY(0, 0), new XY(0, 2), layer, "Wall");
AssertContains(new XY(0, 0), new XY(2, 0), layer, "Wall");
AssertContains(new XY(2, 0), new XY(2, 2), layer, "Wall");
AssertContains(new XY(0, 2), new XY(2, 2), layer, "Wall");
}
private void AddPortals(LayerBuilder builder)
{
for (var row = 0; row < 3; row++)
{
for (var column = 0; column < 3; column++)
{
builder.Put(column, row, new FacilityPortal {
ObjectLayer = ObjectLayer.LowerObject, Endpoint1 = SpecialLocation.OffOfMap, Endpoint2 = new XYZ(column, row, 0)
});
}
}
}
/// <summary>
/// View the legend for the selected layer.
/// </summary>
/// <param name="sender"></param>
/// <param name="e"></param>
private void viewLegendToolStripMenuItem_Click(object sender, EventArgs e)
{
LayerBuilder oBuilder = m_oCurrServerLayers[c_lvLayers.SelectedIndices[0]];
string[] aLegends = oBuilder.GetLegendURLs();
foreach (string szLegend in aLegends)
{
if (!String.IsNullOrEmpty(szLegend))
{
MainForm.BrowseTo(szLegend);
}
}
}
public void LayerBuilder_PutUpperObject_ObjectPutCorrectly()
{
var builder = new LayerBuilder(3, 3);
var painting = new FacilityObject {
Type = "Painting", Orientation = Orientation.Right, ObjectLayer = ObjectLayer.UpperObject
};
builder.PutFloor(1, 1);
builder.Put(0, 1, painting);
var layer = builder.Build();
Assert.IsTrue(layer[0, 1].Contains(painting));
Assert.AreEqual(painting, layer[0, 1].UpperObject);
}
public void LayerBuilder_PutLowerObject_ObjectPutCorrectly()
{
var builder = new LayerBuilder(3, 3);
var cash = new FacilityObject {
Type = "Cash", Orientation = Orientation.Up, ObjectLayer = ObjectLayer.LowerObject
};
builder.PutFloor(1, 1);
builder.Put(1, 1, cash);
var layer = builder.Build();
Assert.IsTrue(layer[1, 1].Contains(cash));
Assert.AreEqual(cash, layer[1, 1].LowerObject);
}
private static long[] AddCnnLayer(IAllocator allocator, SeedSource seedSource, DType elementType, Sequential model, long[] inputSizes, int nOutputPlane, bool useCudnn)
{
var conv = LayerBuilder.BuildConvLayer(allocator, seedSource, elementType, (int)inputSizes[0], (int)inputSizes[3], (int)inputSizes[2], (int)inputSizes[1], nOutputPlane,
new ConvolutionDesc2d(5, 5, 1, 1, 0, 0), useCudnn);
model.Add(conv);
var cdPool = new ConvolutionDesc2d(2, 2, 1, 1, 0, 0);
var poolLayer = LayerBuilder.BuildPoolLayer(allocator, elementType, conv.OutputSizes, cdPool, useCudnn);
model.Add(poolLayer);
model.Add(new ReLULayer(allocator, elementType, poolLayer.OutputSizes));
return(poolLayer.OutputSizes);
}
// Constructs a network with two fully-connected layers; one sigmoid, one softmax
public static void BuildMLPSoftmax(IAllocator allocator, SeedSource seedSource, int batchSize, bool useCudnn, out Sequential model, out ICriterion criterion, out bool outputIsClassIndices)
{
int inputSize = MnistParser.ImageSize * MnistParser.ImageSize;
int hiddenSize = 100;
int outputSize = MnistParser.LabelCount;
var elementType = DType.Float32;
model = new Sequential();
model.Add(new ViewLayer(batchSize, inputSize));
model.Add(new LinearLayer(allocator, seedSource, elementType, inputSize, hiddenSize, batchSize));
model.Add(new SigmoidLayer(allocator, elementType, batchSize, hiddenSize));
model.Add(new LinearLayer(allocator, seedSource, elementType, hiddenSize, outputSize, batchSize));
model.Add(LayerBuilder.BuildLogSoftMax(allocator, elementType, batchSize, outputSize, useCudnn));
criterion = new ClassNLLCriterion(allocator, batchSize, outputSize);
outputIsClassIndices = true; // output of criterion is class indices
}
/// <summary>
/// Turns a list of LayerUris into a list of LayerBuilders.
/// </summary>
/// <param name="oUris"></param>
/// <returns></returns>
private List <LayerBuilder> CreateLayerBuilders(List <LayerUri> oUris)
{
List <LayerBuilder> result = new List <LayerBuilder>();
foreach (LayerUri oUri in oUris)
{
LayerBuilder oLayerToAdd = null;
try
{
oLayerToAdd = oUri.getBuilder(m_oModel);
if (oLayerToAdd != null)
{
result.Add(oLayerToAdd);
}
}
catch (Exception ex)
{
Program.ShowMessageBox(ex.Message, "Dataset Could Not Be Added", MessageBoxButtons.OK, MessageBoxDefaultButton.Button1, MessageBoxIcon.Error);
}
}
return(result);
}
public void Rebuild()
{
LayerBuilder.PopulateLayers(ref _layers);
}
public unsafe static void MNISTExample()
{
//Hyperparameters
Hyperparameters.LearningRate = 0.001f;
Hyperparameters.Optimizer = new SGD();
//Model Creation
var x = new Input(784);
//var dropout = new Dropout(x, 0.1f);
//var model = LayerBuilder.Dense(500, x, "relu");
var model = LayerBuilder.Dense(100, x, "relu");
model = LayerBuilder.Dense(400, model, "relu");
model = LayerBuilder.Dense(200, model, "relu");
model = LayerBuilder.Dense(100, model, "relu");
model = LayerBuilder.Dense(10, model, "softmax");
//Loss Function Creation
var y = new Input(10);
var loss = LayerBuilder.SquaredError(model, y);
//Data preparation
(float[,] traindata, float[,] labels) = LoadMNISTDataSet();
int mnistsize = 42000;
Tensor x_train = Tensor.LoadArrayToDisposedTensor(traindata, new Shape(mnistsize, 784), DeviceConfig.Host_Float);
Tensor y_train = Tensor.LoadArrayToDisposedTensor(labels, new Shape(mnistsize, 10), DeviceConfig.Host_Float);
//Training
int batchsize = 100;
int trainl = 41000;
Stopwatch s = new Stopwatch();
for (int epoch = 0; epoch < 35; epoch++)
{
float l = 0;
float val = 0;
s.Restart();
Console.WriteLine("Epoch " + epoch + " başladı.");
for (int batch = 0; batch < trainl / batchsize; batch++)
{
Tensor batchx = Tensor.Cut(x_train, batch * (batchsize * 784), new Shape(1, batchsize, 784));
Tensor batchy = Tensor.Cut(y_train, batch * (batchsize * 10), new Shape(1, batchsize, 10));
x.SetInput(batchx);
y.SetInput(batchy);
loss.Minimize();
Index zero = new Index(loss.OuterShape);
zero.SetZero();
Tensor res = loss.GetTerm(zero).GetResult();
float *pp = (float *)res.Array;
for (int i = 0; i < res.Shape.TotalSize; i++)
{
l += pp[i];
}
}
for (int batch = trainl / batchsize; batch < mnistsize / batchsize; batch++)
{
Tensor batchx = Tensor.Cut(x_train, batch * (batchsize * 784), new Shape(1, batchsize, 784));
Tensor batchy = Tensor.Cut(y_train, batch * (batchsize * 10), new Shape(1, batchsize, 10));
model.DeleteTerms();
x.SetInput(batchx);
y.SetInput(batchy);
Index zero = new Index(model.OuterShape);
zero.SetZero();
model.PreCheck();
Tensor res = model.GetTerm(zero).GetResult();
for (int i = 0; i < batchsize; i++)
{
int myans = MaxId((float *)res.Array + i * 10);
int correctres = MaxId((float *)batchy.Array + i * 10);
val += (myans == correctres ? 1 : 0);
}
}
s.Stop();
Console.WriteLine("Epoch " + epoch + " biti.");
Console.WriteLine("Loss: " + l / trainl);
Console.WriteLine("Validation: " + val / (mnistsize - trainl));
Console.WriteLine("Time: " + s.ElapsedMilliseconds + "ms");
}
PrintPools();
while (true)
{
try
{
float[] data = LoadCurrentImage();
Tensor x_test = Tensor.LoadArrayToDisposedTensor(data, new Shape(1, 1, 784), DeviceConfig.Host_Float);
model.DeleteTerms();
x.SetInput(x_test);
Index zero = new Index(model.OuterShape);
zero.SetZero();
model.PreCheck();
Tensor res = model.GetTerm(zero).GetResult();
Console.WriteLine("Result: " + res);
Console.WriteLine("Digit Prediction: " + MaxId((float *)res.Array));
Console.WriteLine("-----------");
}
catch (Exception)
{
}
Thread.Sleep(500);
}
}
public FacilityMap(IWorld world, MapInstruction inst)
{
_world = world;
inst.Layers.ForEach(x => Add(LayerBuilder.Assemble(x)));
inst.Portals.ForEach(x => this[x.Location].Put(PortalFactory.Create(x)));
}
static void Main(string[] args)
{
var rd = new Random();
var values = new[] { .1f, .002f };
var watch = System.Diagnostics.Stopwatch.StartNew();
ProcessingDevice.Device = Device.CPU;
watch.Stop();
Console.WriteLine($"Device Time: {watch.ElapsedMilliseconds}ms");
var hiddens = new LayerBuilder(2, 4, values[0])
.Supervised()
.WithLeakRelu()
.Hidden()
.WithSGD()
.WithMomentum(values[1])
.FullSynapse()
.Build();
var hiddens2 = new LayerBuilder(2, 2, values[0])
.Supervised()
.WithLeakRelu()
.Hidden()
.WithSGD()
.WithMomentum(values[1])
.FullSynapse()
.Build();
var outputs = new LayerBuilder(2, 2, values[0])
.Supervised()
.WithSigmoid()
.Output()
.WithSGD()
.WithMomentum(values[1])
.FullSynapse()
.Build();
var loss = new SquareLossFunction();
watch = System.Diagnostics.Stopwatch.StartNew();
watch.Stop();
Console.WriteLine($"Sinapse Time: {watch.ElapsedMilliseconds}ms");
var trainingValues = new[]
{
new[] { 0f, 0f, 0f, 0f },
new[] { 1f, 0f, 0f, 0f },
new[] { 0f, 1f, 0f, 0f },
new[] { 0f, 0f, 1f, 0f },
new[] { 0f, 0f, 0f, 1f },
new[] { 1f, 0f, 0f, 0f },
new[] { 1f, 1f, 0f, 0f },
new[] { 0f, 1f, 1f, 0f },
new[] { 0f, 0f, 1f, 1f },
new[] { 0f, 0f, 0f, 1f },
new[] { 0f, 1f, 0f, 0f },
new[] { 1f, 0f, 1f, 0f },
new[] { 0f, 1f, 0f, 1f },
new[] { 1f, 0f, 1f, 0f },
new[] { 0f, 1f, 0f, 1f },
new[] { 0f, 0f, 1f, 0f },
new[] { 1f, 0f, 0f, 1f },
new[] { 1f, 1f, 0f, 0f },
new[] { 0f, 1f, 1f, 0f },
new[] { 0f, 0f, 1f, 1f },
new[] { 1f, 1f, 1f, 0f },
new[] { 1f, 1f, 1f, 1f }
};
var desiredValues = new[]
{
new[] { 1f, 0f },
new[] { 1f, 0f },
new[] { 1f, 0f },
new[] { 1f, 0f },
new[] { 0f, 1f },
new[] { 1f, 0f },
new[] { 1f, 0f },
new[] { 1f, 0f },
new[] { 0f, 1f },
new[] { 0f, 1f },
new[] { 1f, 0f },
new[] { 1f, 0f },
new[] { 0f, 1f },
new[] { 1f, 0f },
new[] { 0f, 1f },
new[] { 1f, 0f },
new[] { 0f, 1f },
new[] { 1f, 0f },
new[] { 1f, 0f },
new[] { 0f, 1f },
new[] { 1f, 0f },
new[] { 0f, 1f }
};
int cont = 0;
int sizeTrain = 10;
var e = double.MaxValue;
while (true)
{
watch = System.Diagnostics.Stopwatch.StartNew();
e = 0;
for (int i = 0; i < sizeTrain; i++)
{
var index = rd.Next(0, trainingValues.Length);
var inputs = trainingValues[index];
var desireds = desiredValues[index];
//watch = System.Diagnostics.Stopwatch.StartNew();
// Feed Forward
var _h = hiddens.Output(inputs);
var _h2 = hiddens2.Output(_h);
var _o = outputs.Output(_h2);
//watch.Stop();
//Console.WriteLine($"\nForward Time: { watch.ElapsedMilliseconds}ms");
//Thread.Sleep(100);
//watch = System.Diagnostics.Stopwatch.StartNew();
// Backward
var _oe = ((ISupervisedLearning)outputs).Learn(_h2, desireds);
var _he2 = ((ISupervisedLearning)hiddens2).Learn(_h, _oe);
((ISupervisedLearning)hiddens).Learn(inputs, _he2);
//watch.Stop();
//Console.WriteLine($"\nBackward Time: { watch.ElapsedMilliseconds}ms");
// Error
var e0 = Math.Abs(_o[0] - desireds[0]);
var e1 = Math.Abs(_o[1] - desireds[1]);
var error = Math.Sqrt(Math.Abs(e0 * e0 + e1 * e0));
e += error / 2.0;
}
e /= sizeTrain;
cont++;
watch.Stop();
var time = watch.ElapsedMilliseconds;
Console.WriteLine($"Interactions: {cont}\nError: {e}");
//Console.WriteLine($"Interactions: {cont}\nError: {e}\nTime: {time / (double)sizeTrain}ms");
Console.Title =
$"TSPS (Training Sample per Second): {Math.Ceiling(1000d / ((double) time / (double)sizeTrain))}";
}
}
static void Main(string[] args)
{
var events = ReadEventFile("trainingEvents.csv");
var ins = events
.Where(e => (e.Open - e.NextLow) / e.Open > .1f)
.Select(evts => evts.GetInputArray());
Console.WriteLine($"Qualified Events: {ins.Count()}");
var unsupervisedTests = ins.Select(i => Tuple.Create(i, i));
var supervisedTests = events.Select(evt => Tuple.Create(evt.GetInputArray(), evt.GetOutputArray()));
var builder = new LayerBuilder();
var description = builder.BuildDescription(5, new[]
{
new LayerBuilder.LayerSpec(10, "sum", "softplus"),
new LayerBuilder.LayerSpec(10, "sum", "softplus"),
new LayerBuilder.LayerSpec(10, "sum", "tanh"),
new LayerBuilder.LayerSpec(10, "sum", "softplus"),
new LayerBuilder.LayerSpec(10, "sum", "softplus"),
new LayerBuilder.LayerSpec(5, "sum", null)
});
//var description = builder.BuildDescription(5, new[]
//{
// new LayerBuilder.LayerSpec(5, "sum", "softplus"),
// new LayerBuilder.LayerSpec(6, "sum", "tanh"),
// new LayerBuilder.LayerSpec(4, "sum", "softplus"),
// new LayerBuilder.LayerSpec(5, "sum", null)
//});
var net = Net.FromDescription(description);
var trainer = new SimpleTrainer();
trainer.Train(
net: net,
tests: unsupervisedTests,
desiredError: 5e-6f,
maxEpochs: 200,
learningRate: 0.75f);
trainer.Train(
net: net,
tests: unsupervisedTests,
desiredError: 5e-6f,
maxEpochs: 9500,
learningRate: 0.5f);
trainer.Train(
net: net,
tests: unsupervisedTests,
desiredError: 1e-8f,
maxEpochs: 200,
learningRate: .25f);
trainer.Train(
net: net,
tests: unsupervisedTests,
desiredError: 1e-8f,
maxEpochs: 200,
learningRate: .125f);
trainer.Train(
net: net,
tests: unsupervisedTests,
desiredError: 1e-8f,
maxEpochs: 200,
learningRate: .0625f);
for (var i = 0; i < 5; i++)
{
var nextDescription = net.Description;
var firstSigmoidId = nextDescription.Nodes.First(n => n.Processor == "tanh").NodeId;
nextDescription.Nodes = nextDescription.Nodes.Where(n => n.NodeId != firstSigmoidId).ToArray();
foreach (var node in nextDescription.Nodes)
{
node.Inputs = node.Inputs
.Where(inp => inp.FromInputVector || inp.InputId != firstSigmoidId)
.ToArray();
}
net = Net.FromDescription(nextDescription);
Console.WriteLine($"Removed {i + 1} sigmoids");
trainer.Train(
net: net,
tests: unsupervisedTests,
desiredError: 1e-6f,
maxEpochs: 400,
learningRate: 0.5f);
}
trainer.Train(
net: net,
tests: unsupervisedTests,
desiredError: 1e-6f,
maxEpochs: 500,
learningRate: 0.5f);
trainer.Train(
net: net,
tests: unsupervisedTests,
desiredError: 1e-6f,
maxEpochs: 400,
learningRate: 0.25f);
var finalDescription = net.Description;
var outsRemoved = new[] { finalDescription.Outputs[1], finalDescription.Outputs[2], finalDescription.Outputs[3], finalDescription.Outputs[4] };
finalDescription.Nodes = finalDescription.Nodes
.Where(n => !outsRemoved.Contains(n.NodeId))
.ToArray();
finalDescription.Outputs = new[] { finalDescription.Outputs[0] };
var nextNet = Net.FromDescription(finalDescription);
trainer.Train(
net: nextNet,
tests: supervisedTests,
desiredError: 1e-8f,
maxEpochs: 20000,
learningRate: .25f);
trainer.Train(
net: nextNet,
tests: supervisedTests,
desiredError: 1e-8f,
maxEpochs: 20000,
learningRate: .125f);
trainer.Train(
net: nextNet,
tests: supervisedTests,
desiredError: 1e-8f,
maxEpochs: 2000,
learningRate: .0625f);
trainer.Train(
net: nextNet,
tests: supervisedTests,
desiredError: 1e-8f,
maxEpochs: 2000,
learningRate: .03125f);
var final = net.Description;
var finalText = JsonConvert.SerializeObject(final);
using (var writer = File.CreateText("out3.json"))
{
writer.Write(finalText);
}
Console.WriteLine();
Console.ReadLine();
}
public void TestReadAndFillDoesntChangeError()
{
var tests = new[]
{
new[] { new[] { 1f, 1f }, new[] { -1f } },
new[] { new[] { 1f, -1f }, new[] { 1f } },
new[] { new[] { -1f, 1f }, new[] { 1f } },
new[] { new[] { -1f, -1f }, new[] { -1f } }
};
var builder = new LayerBuilder();
var desc = builder.BuildDescription(2, new[]
{
new LayerBuilder.LayerSpec(2, "sum", "tanh"),
new LayerBuilder.LayerSpec(1, "sum", "tanh")
});
var net = Net.FromDescription(desc);
var train = net.GetTrainingFunction();
var weights = new float[net.NumberOfWeights];
var deltas = new float[net.NumberOfWeights];
net.FillWeights(weights);
var loss = 0f;
for (var i = 0; i < 15001; i++)
{
loss = 0;
foreach (var test in tests)
{
loss += train(test[0], test[1], weights, deltas);
for (var j = 0; j < weights.Length; j++)
{
weights[j] -= deltas[j] / 2f;
}
}
if (i % 300 == 0)
{
Console.WriteLine($"{i}, {loss}");
}
}
loss = 0;
foreach (var test in tests)
{
loss += train(test[0], test[1], weights, deltas);
}
Console.WriteLine(loss);
Console.WriteLine(string.Join(", ", weights));
net.ReadWeights(weights);
var newWeights = new float[net.NumberOfWeights];
net.FillWeights(newWeights);
Console.WriteLine(string.Join(", ", newWeights));
for (var i = 0; i < weights.Length; i++)
{
Assert.AreEqual(weights[i], newWeights[i]);
}
var newLoss = 0f;
foreach (var test in tests)
{
newLoss += train(test[0], test[1], newWeights, new float[net.NumberOfWeights]);
}
Console.WriteLine(newLoss);
Assert.AreEqual(loss, newLoss);
}
public unsafe static void XORExample()
{
//Hyperparameters
Hyperparameters.LearningRate = 0.1f;
Hyperparameters.Optimizer = new SGD();
//Model Creation
var l1 = LayerBuilder.Dense(16, "sigmoid");
var l2 = LayerBuilder.Dense(1, "sigmoid")[l1];
var x = new Input(2);
Layer model = l2[x];
//Loss Function Creation
var y = new Input(1);
var loss = LayerBuilder.SquaredError(model, y);
//Data preparation
Tensor x_train = new Tensor((1, 4, 2), DeviceConfig.Host_Float);
Tensor y_train = new Tensor((1, 4, 1), DeviceConfig.Host_Float);
float *xt = (float *)x_train.Array;
float *yt = (float *)y_train.Array;
// 1,1 = 0
// 1,0 = 1
// 0,1 = 1
// 0,0 = 0
xt[0] = 1; xt[1] = 1;
xt[2] = 1; xt[3] = 0;
xt[4] = 0; xt[5] = 1;
xt[6] = 0; xt[7] = 0;
yt[0] = 0;
yt[1] = 1;
yt[2] = 1;
yt[3] = 0;
//Give data to the model
x.SetInput(x_train);
y.SetInput(y_train);
Stopwatch s = new Stopwatch();
s.Start();
//Minimizing
loss.PreCheck();
Index a = new Index(model.OuterShape);
a.SetZero();
for (int epoch = 0; epoch < 100000; epoch++)
{
loss.Minimize();
if (epoch % 5000 == 0)
{
float res = ((float *)loss.GetTerm(a).GetResult().Array)[0];
res += ((float *)loss.GetTerm(a).GetResult().Array)[1];
res += ((float *)loss.GetTerm(a).GetResult().Array)[2];
res += ((float *)loss.GetTerm(a).GetResult().Array)[3];
Console.WriteLine(res);
}
}
s.Stop();
Console.WriteLine("Time Elapsed: " + s.ElapsedMilliseconds);
//Print Pools
PrintPools();
//Print the results
var result = model.GetTerm(a).GetResult();
Console.WriteLine("Results: " + result);
//Print the results of clone model
Input x2 = new Input(2);
x2.SetInput(x_train);
var clonemodel = l2[x2];
clonemodel.PreCheck();
var result2 = clonemodel.GetTerm(a).GetResult();
Console.WriteLine("Results: " + result2);
clonemodel.DeleteTerms();
model.DeleteTerms();
}
/// <summary>
/// Basic initialising neural network, before initialising fully use Initialise() method
/// </summary>
/// <param name="hidenLayersNeuronCount">int array that contains number of neurons for each hiden layer</param>
/// <param name="inputCount">number of input neurons</param>
/// <param name="outputCount">number of output neurons</param>
public NeuralNetwork(int[] hidenLayersNeuronCount, int inputCount, int outputCount)
{
layers = LayerBuilder.CreateNewNeuralNetwork(hidenLayersNeuronCount, inputCount, outputCount);
}
