public void SimulationRunsViaRunner()
{
using (var runner = new MyProjectRunner())
{
MyProject project = runner.CreateProject(typeof(MyTestingWorld));
var node = project.CreateNode <MyCSharpNode>();
project.Network.AddChild(node);
runner.RunAndPause(1);
}
}
public void CreatesAndRunsMNIST()
{
using (var runner = new MyProjectRunner())
{
MyProject project = runner.CreateProject(typeof(MNISTWorld));
var world = project.World as MNISTWorld;
var neuralGroup = project.CreateNode <MyNeuralNetworkGroup>();
project.Network.AddChild(neuralGroup);
var hiddenLayer = project.CreateNode <MyHiddenLayer>();
neuralGroup.AddChild(hiddenLayer);
var outputLayer = project.CreateNode <MyOutputLayer>();
neuralGroup.AddChild(outputLayer);
var accumulator = project.CreateNode <MyAccumulator>();
neuralGroup.AddChild(accumulator);
// Connect the nodes.
project.Connect(project.Network.GroupInputNodes[0], neuralGroup, 0, 0);
project.Connect(project.Network.GroupInputNodes[1], neuralGroup, 0, 1);
project.Connect(neuralGroup.GroupInputNodes[0], hiddenLayer, 0, 0);
project.Connect(neuralGroup.GroupInputNodes[1], outputLayer, 0, 1);
project.Connect(hiddenLayer, outputLayer, 0, 0);
project.Connect(outputLayer, accumulator, 1, 0);
// Setup the nodes.
var sendMnistData = world.SendMNISTTrainData;
Assert.NotNull(sendMnistData);
sendMnistData.ExpositionTime = 1;
world.BitmapOrder = ExampleOrderOption.Shuffle;
world.OneHot = true;
hiddenLayer.Neurons = 40;
accumulator.ApproachValue.ApproachMethod = MyAccumulator.MyApproachValueTask.SequenceType.Momentum;
accumulator.ApproachValue.Delta = 0.1f;
accumulator.ApproachValue.Target = 0;
accumulator.ApproachValue.Factor = 0.9f;
// Enable tasks.
project.World.EnableDefaultTasks();
neuralGroup.EnableDefaultTasks();
neuralGroup.RMS.Enabled = true;
hiddenLayer.EnableDefaultTasks();
outputLayer.EnableDefaultTasks();
accumulator.EnableDefaultTasks();
accumulator.ApproachValue.Enabled = true;
// Run the simulation.
runner.RunAndPause(100);
float error = runner.GetValues(accumulator.Id)[0];
Assert.True(error < 0.5f);
//runner.SaveProject(@"c:\foobar.brain");
}
}
public void CreatesAndRunsMNIST()
{
using (var runner = new MyProjectRunner())
{
MyProject project = runner.CreateProject(typeof(MyMNISTWorld));
MyWorld world = project.World;
var neuralGroup = project.CreateNode <MyNeuralNetworkGroup>();
project.Network.AddChild(neuralGroup);
var hiddenLayer = project.CreateNode <MyHiddenLayer>();
neuralGroup.AddChild(hiddenLayer);
var outputLayer = project.CreateNode <MyOutputLayer>();
neuralGroup.AddChild(outputLayer);
var accumulator = project.CreateNode <MyAccumulator>();
neuralGroup.AddChild(accumulator);
// Connect the nodes.
project.Connect(project.Network.GroupInputNodes[0], neuralGroup, 0, 0);
project.Connect(project.Network.GroupInputNodes[1], neuralGroup, 0, 1);
project.Connect(neuralGroup.GroupInputNodes[0], hiddenLayer, 0, 0);
project.Connect(neuralGroup.GroupInputNodes[1], outputLayer, 0, 1);
project.Connect(hiddenLayer, outputLayer, 0, 0);
project.Connect(outputLayer, accumulator, 1, 0);
// Setup the nodes.
MyTask sendMnistData = world.GetTaskByPropertyName("SendTrainingMNISTData");
Assert.NotNull(sendMnistData);
sendMnistData.GetType().GetProperty("RandomEnumerate").SetValue(sendMnistData, true);
sendMnistData.GetType().GetProperty("ExpositionTime").SetValue(sendMnistData, 1);
world.GetType().GetProperty("Binary").SetValue(world, true);
hiddenLayer.Neurons = 40;
accumulator.ApproachValue.ApproachMethod = MyAccumulator.MyApproachValueTask.SequenceType.Momentum;
accumulator.ApproachValue.Delta = 0.1f;
accumulator.ApproachValue.Target = 0;
accumulator.ApproachValue.Factor = 0.9f;
// Enable tasks.
project.World.EnableDefaultTasks();
neuralGroup.EnableDefaultTasks();
neuralGroup.RMS.Enabled = true;
hiddenLayer.EnableDefaultTasks();
outputLayer.EnableDefaultTasks();
accumulator.EnableDefaultTasks();
accumulator.ApproachValue.Enabled = true;
// Run the simulation.
runner.RunAndPause(100);
float error = runner.GetValues(accumulator.Id)[0];
Assert.True(error < 0.5f);
//runner.SaveProject(@"c:\foobar.brain");
}
}