//Projects corpus
public SymbolCorpus project(SymbolCorpus control)
{
//Create same dimension output Corpus
SymbolCorpus outputCorpus = new SymbolCorpus();
outputCorpus.dialogues = new SymbolDialogue[control.dialogues.Length];
//Process
SymbolDialogue controlDialogue;
SymbolSentence controlSentence;
for (int i = 0; i < control.dialogues.Length; i++)
{
controlDialogue = control.dialogues[i];
//Create same dimension output dialogue
outputCorpus.dialogues[i] = new SymbolDialogue();
outputCorpus.dialogues[i].sentences = new SymbolSentence[control.dialogues[i].sentences.Length];
for (int j = 0; j < controlDialogue.sentences.Length; j++)
{
controlSentence = controlDialogue.sentences[j];
outputCorpus.dialogues[i].sentences[j] = new SymbolSentence();
//project corpus, reset state per sentence
state = 0;
outputCorpus.dialogues[i].sentences[j].symbols = project(controlSentence.symbols);
}
}
return(outputCorpus);
}
public void train(SymbolCorpus trainSet, int cycles)
{
Random rdm = new Random();
AutoStatePredictor predictor = new AutoStatePredictor(symbolSize, stateSize);
//Make copy of current
copyInto(this, predictor);
if (predictor.trainingDepth == 0)
{
randomOverride(predictor, rdm);
//Read and predict analysis
predictor.testPredict(trainSet);
//TODO Generate some more randoms and choose best - might not be needed since leap Intensity is so high at the beginning
}
//Initialize training variables
AutoStatePredictor testTable = new AutoStatePredictor(predictor.symbolSize, predictor.stateSize);
AutoStatePredictor temp;
double leapIntensity = predictor.lastLeapIntensity * 2;
if (leapIntensity > 0.9)
{
leapIntensity = 0.9;
}
for (int i = 0; i < cycles; i++)
{
//Mutation
leapOverride(testTable, predictor, rdm, leapIntensity);
//Read and predict analysis
testTable.testPredict(trainSet);
if (testTable.accuracy > predictor.accuracy)
{
//Best predictor found
//Soft change with memory for sucessful mutations intensity (x2 because of random average behaviour)
leapIntensity = 0.8 * leapIntensity + 0.2 * (2 * testTable.lastLeapIntensity);
if (leapIntensity > 0.9)
{
leapIntensity = 0.9; //maximum allowed mutation intensity
}
//Swap predictors for memory usage
temp = predictor;
predictor = testTable;
testTable = temp;
}
}
copyInto(predictor, this);
}
private void bake(SymbolCorpus trainSet)
{
SymbolDialogue dialogue;
SymbolSentence sentence;
byte feed = 0;
for (int i = 0; i < trainSet.dialogues.Length; i++)
{
dialogue = trainSet.dialogues[i];
//Reset states
if (deepLayers != null)
{
for (int k = 0; k < deepLayers.Length; k++)
{
deepLayers[k].state = 0;
}
}
outputLayer.state = 0;
for (int j = 0; j < dialogue.sentences.Length; j++)
{
sentence = dialogue.sentences[j];
//Predict first symbol
feed = 0;
if (deepLayers != null)
{
for (int k = 0; k < deepLayers.Length; k++)
{
feed = deepLayers[k].process(feed);
}
}
outputLayer.bake(feed, sentence.symbols[0]);
//Predict remaining symbols
for (int l = 1; l < sentence.symbols.Length; l++)
{
feed = sentence.symbols[l - 1];
if (deepLayers != null)
{
for (int k = 0; k < deepLayers.Length; k++)
{
feed = deepLayers[k].process(feed);
}
}
outputLayer.bake(feed, sentence.symbols[l]);
}
}
}
outputLayer.finalizeBake();
}
public SymbolCorpus truncanteDialogues(int dialogueCount)
{
SymbolCorpus truncated = new SymbolCorpus();
truncated.dialogues = new SymbolDialogue[dialogueCount];
for (int i = 0; i < truncated.dialogues.Length; i++)
{
truncated.dialogues[i] = dialogues[i];
}
return(truncated);
}
public SymbolCorpus textToSymbol(Corpus text)
{
SymbolCorpus symbols = new SymbolCorpus();
symbols.dialogues = new SymbolDialogue[text.dialogues.Length];
for (int i = 0; i < symbols.dialogues.Length; i++)
{
symbols.dialogues[i] = textToSymbol(text.dialogues[i]);
}
return(symbols);
}
public Corpus symbolToText(SymbolCorpus symbols)
{
Corpus text = new Corpus();
text.dialogues = new Corpus.Dialogue[symbols.dialogues.Length];
for (int i = 0; i < text.dialogues.Length; i++)
{
text.dialogues[i] = symbolToText(symbols.dialogues[i]);
}
return(text);
}
public CorpusResult process(SymbolCorpus input)
{
CorpusResult result = new CorpusResult();
//Create same dimension process Corpus
result.states = new SymbolCorpus();
result.states.dialogues = new SymbolDialogue[input.dialogues.Length];
result.hits = new SymbolCorpus();
result.hits.dialogues = new SymbolDialogue[input.dialogues.Length];
result.predictions = new SymbolCorpus();
result.predictions.dialogues = new SymbolDialogue[input.dialogues.Length];
//Process
SymbolDialogue inputDialogue;
SymbolSentence inputSentence;
ProcessResult sentenceResult;
for (int i = 0; i < input.dialogues.Length; i++)
{
inputDialogue = input.dialogues[i];
//Create same dimension process dialogue
result.states.dialogues[i] = new SymbolDialogue();
result.states.dialogues[i].sentences = new SymbolSentence[input.dialogues[i].sentences.Length];
result.hits.dialogues[i] = new SymbolDialogue();
result.hits.dialogues[i].sentences = new SymbolSentence[input.dialogues[i].sentences.Length];
result.predictions.dialogues[i] = new SymbolDialogue();
result.predictions.dialogues[i].sentences = new SymbolSentence[input.dialogues[i].sentences.Length];
for (int j = 0; j < inputDialogue.sentences.Length; j++)
{
inputSentence = inputDialogue.sentences[j];
result.states.dialogues[i].sentences[j] = new SymbolSentence();
result.hits.dialogues[i].sentences[j] = new SymbolSentence();
result.predictions.dialogues[i].sentences[j] = new SymbolSentence();
//predict corpus, reset state per sentence
state = 0;
sentenceResult = process(inputSentence.symbols);
result.states.dialogues[i].sentences[j].symbols = sentenceResult.states;
result.hits.dialogues[i].sentences[j].symbols = sentenceResult.hits;
result.predictions.dialogues[i].sentences[j].symbols = sentenceResult.predictions;
}
}
return(result);
}
public void StartTrain(SymbolCorpus trainSet)
{
this.trainSet = trainSet;
int threadCount = 2;
training = true;
threadStatus = new List <string>();
for (int i = 0; i < threadCount; i++)
{
threadStatus.Add("");
}
best = new Couppy(target.layerStateSizes, target.layerOutputSizes, target.translator);
Couppy.copyInto(target, best);
best.outputLayer.reset();
best.bake(trainSet);
TrainingUpdated?.Invoke("Layer training started with: " + best.outputLayer.getStats());
for (int i = 0; i < trainSampleCount.Length; i++)
{
trainSampleCount[i] = 1;
trainAccumulatedAccuracies[i] = best.outputLayer.accuracy;
}
for (int i = 0; i < threadCount; i++)
{
//Add thread lock
readLocks.Add(new object());
trainLocks.Add(new object());
//Create thread resource
Couppy tester = new Couppy(best.layerStateSizes, best.layerOutputSizes, best.translator);
chatBots.Add(tester);
//Create and start thread
Thread tt = new Thread(ThreadTrain);
threads.Add(tt);
}
//Start threads after to avoid first cycle conflicts
for (int i = 0; i < threadCount; i++)
{
threads[i].Start(i);
}
}
//Predicts corpus only to update error
public void testPredict(SymbolCorpus input)
{
//Process
SymbolDialogue inputDialogue;
SymbolSentence inputSentence;
for (int i = 0; i < input.dialogues.Length; i++)
{
inputDialogue = input.dialogues[i];
for (int j = 0; j < inputDialogue.sentences.Length; j++)
{
inputSentence = inputDialogue.sentences[j];
//predict corpus, reset state per sentence
state = 0;
testPredict(inputSentence.symbols);
}
}
}
public void StartTrain(SymbolCorpus trainSet)
{
this.trainSet = trainSet;
int threadCount = 8;
training = true;
best = new HybridStatePredictor(target.symbolSize, target.stateSize);
HybridStatePredictor.copyInto(target, best);
best.reset();
best.testPredict(trainSet);
Console.WriteLine("Layer training started with: " + best.getStats());
for (int i = 0; i < trainSampleCount.Length; i++)
{
trainSampleCount[i] = 1;
trainAccumulatedAccuracies[i] = best.accuracy;
}
for (int i = 0; i < threadCount; i++)
{
//Add thread lock
readLocks.Add(new object());
trainLocks.Add(new object());
//Create thread resource
HybridStatePredictor tester = new HybridStatePredictor(best.symbolSize, best.stateSize);
predictors.Add(tester);
//Create and start thread
Thread tt = new Thread(ThreadTrain);
threads.Add(tt);
}
//Start threads after to avoid first cycle conflicts
for (int i = 0; i < threadCount; i++)
{
threads[i].Start(i);
}
}
public void StartTrain(SymbolCorpus trainSet)
{
trainer = new Trainer(this);
trainer.StartTrain(trainSet);
}
public void train2(SymbolCorpus trainSet, double minutes)
{
Random rdm = new Random();
HybridStatePredictor predictor = new HybridStatePredictor(symbolSize, stateSize);
//Make copy of current
copyInto(this, predictor);
if (predictor.trainingDepth == 0)
{
randomOverride(predictor, rdm);
//Read and predict analysis
predictor.testPredict(trainSet);
//TODO Generate some more randoms and choose best - might not be needed since leap Intensity is so high at the beginning
}
//Initialize training variables
HybridStatePredictor testTable = new HybridStatePredictor(predictor.symbolSize, predictor.stateSize);
HybridStatePredictor temp;
double leapIntensity = predictor.lastLeapIntensity * 2 * 4;
if (leapIntensity > 0.9)
{
leapIntensity = 0.9;
}
Stopwatch stopwatch = new Stopwatch();
stopwatch.Start();
Console.Write("Training:");
int percentil = 1;
while (stopwatch.ElapsedMilliseconds < 1000 * 60 * minutes)
{
if (stopwatch.ElapsedMilliseconds > (1000 * 60 * minutes / 10.0) * percentil)
{
Console.Write("$"); //print training tick
percentil++;
}
//Mutation
//leapOverride(testTable, predictor, rdm, leapIntensity);
//Read and predict analysis
testTable.testPredict(trainSet);
if (testTable.accuracy > predictor.accuracy)
{
//Best predictor found
//Soft change with memory for sucessful mutations intensity (x2 because of random average behaviour)
leapIntensity = 0.8 * leapIntensity + 0.2 * (2 * testTable.lastLeapIntensity * 4);
if (leapIntensity > 0.9)
{
leapIntensity = 0.9; //maximum allowed mutation intensity
}
//Swap predictors for memory usage
temp = predictor;
predictor = testTable;
testTable = temp;
}
}
stopwatch.Stop();
Console.WriteLine();
copyInto(predictor, this);
}