/// <summary>
/// Creates a node and any other connected nodes from a serialised execution graph
/// </summary>
/// <param name="factory"></param>
/// <param name="graph">Serialised graph</param>
public static INode CreateFrom(this GraphFactory factory, Models.ExecutionGraph graph)
{
// create the input node
var nodeTable = new Dictionary <string, INode>();
var ret = factory.Create(graph.InputNode);
nodeTable.Add(ret.Id, ret);
// create the other nodes
foreach (var node in graph.OtherNodes)
{
var n = factory.Create(node);
if (!nodeTable.ContainsKey(n.Id))
{
nodeTable.Add(n.Id, n);
}
}
// let each node know it has been deserialised and access to the entire graph
foreach (var item in nodeTable)
{
item.Value.OnDeserialise(nodeTable);
}
// create the wires between nodes
foreach (var wire in graph.Wires)
{
var from = nodeTable[wire.FromId];
var to = nodeTable[wire.ToId];
from.Output.Add(new WireToNode(to, wire.InputChannel));
}
return(ret);
}
public void TestRecurrent()
{
var data = BinaryIntegers.Addition(100, false).Split(0);
var graph = new GraphFactory(_lap);
var errorMetric = graph.ErrorMetric.BinaryClassification;
graph.CurrentPropertySet
.Use(graph.GradientDescent.Adam)
.Use(graph.GaussianWeightInitialisation(false, 0.1f, GaussianVarianceCalibration.SquareRoot2N));
// create the engine
var trainingData = graph.CreateDataSource(data.Training);
var testData = trainingData.CloneWith(data.Test);
var engine = graph.CreateTrainingEngine(trainingData, learningRate: 0.01f, batchSize: 16);
// build the network
const int HIDDEN_LAYER_SIZE = 32, TRAINING_ITERATIONS = 5;
var memory = new float[HIDDEN_LAYER_SIZE];
var network = graph.Connect(engine)
.AddSimpleRecurrent(graph.ReluActivation(), memory)
.AddFeedForward(engine.DataSource.OutputSize)
.Add(graph.ReluActivation())
.AddBackpropagationThroughTime(errorMetric)
;
// train the network for twenty iterations, saving the model on each improvement
BrightWire.Models.ExecutionGraph bestGraph = null;
engine.Train(TRAINING_ITERATIONS, testData, errorMetric, bn => bestGraph = bn.Graph);
// export the graph and verify it against some unseen integers on the best model
var executionEngine = graph.CreateEngine(bestGraph ?? engine.Graph);
var testData2 = graph.CreateDataSource(BinaryIntegers.Addition(8, true));
var results = executionEngine.Execute(testData2);
}
public static ExecutionGraph TrainModel(IGraphTrainingEngine engine, NetworkConfig config, DataSet dataset, string outputModelPath)
{
BrightWire.Models.ExecutionGraph bestGraph = null;
engine.Train(config.TRAINING_ITERATIONS, dataset.TestData, config.ERROR_METRIC, model => {
bestGraph = model.Graph;
if (!String.IsNullOrWhiteSpace(outputModelPath))
{
using (var file = new FileStream(outputModelPath, FileMode.Create, FileAccess.Write))
{
Serializer.Serialize(file, model);
}
}
});
return(bestGraph);
}
