//public static MultiLayeredPerceptronClassifier RESET
//{
// get
// {
// lock (padlock)
// {
// instance = new MultiLayeredPerceptronClassifier(64, 10, 3, 52);
// instance.isTrained = false;
// return instance;
// }
// }
//}
/*
* Load a MLP from a saved state from file.
*/
public static void CreateFromFile(MLPState mstate)
{
lock (padlock)
{
if (instance == null)
{
instance = new MultiLayeredPerceptronClassifier(64, 10, 3, 52);
instance.inputNeuronsArr = mstate.inputNeuronsArr;
instance.hiddenNeuronsArr = mstate.hiddenNeuronsArr;
instance.outputNeuronsArr = mstate.outputNeuronsArr;
instance.weightsArr = mstate.weightsArr;
instance.tempWeightsArr = mstate.tempWeightsArr;
instance.prevWeightsArr = mstate.prevWeightsArr;
instance.inputs = mstate.inputs;
instance.outputs = mstate.outputs;
instance.hiddenLayers = mstate.hiddenLayers;
instance.hiddenNeurons = mstate.hiddenNeurons;
instance.meanSqErr = mstate.meanSqErr;
instance.epochs = mstate.epochs;
instance.error = mstate.error;
instance.isTrained = true;
}
}
}
private void saveMLPState(MultiLayeredPerceptronClassifier mlp, float accuracy)
{
MLPState mstate = new MLPState()
{
inputNeuronsArr = mlp.inputNeuronsArr,
hiddenNeuronsArr = mlp.hiddenNeuronsArr,
outputNeuronsArr = mlp.outputNeuronsArr,
weightsArr = mlp.weightsArr,
tempWeightsArr = mlp.tempWeightsArr,
prevWeightsArr = mlp.prevWeightsArr,
inputs = mlp.inputs,
outputs = mlp.outputs,
hiddenLayers = mlp.hiddenLayers,
hiddenNeurons = mlp.hiddenNeurons,
meanSqErr = mlp.meanSqErr,
epochs = mlp.epochs,
error = mlp.error
};
// write the data (overwrites)
using (var stream = new StreamWriter(MLP_STATE_FILE, append: false)) /// + "_" + accuracy
{
stream.Write(JsonConvert.SerializeObject(mstate));
}
}
private void loadMLPState()
{
using (var stream = new StreamReader(MLP_STATE_FILE))
{
MLPState mstate = JsonConvert.DeserializeObject <MLPState>(stream.ReadToEnd());
// Load state from file.
MultiLayeredPerceptronClassifier.CreateFromFile(mstate);
}
}
