- moved numEpochs and miniBatchSize to RBM::Params

git-svn-id: http://moon:8086/svn/software/trunk/projects/Rbm@606 b431acfa-c32f-4a4a-93f1-934dc6c82436

source/Rbm.cpp

@@ -57,7 +57,7 @@ Json::Value Rbm::toJson() const
 	return rbm;
 }
 
-void Rbm::train(const arma::mat& batch, size_t miniBatchSize, size_t numEpochs, IListener* pListener)
+void Rbm::train(const arma::mat& batch, IListener* pListener)
 {
 	Status status;
 	size_t epoch;
@@ -83,7 +83,7 @@ void Rbm::train(const arma::mat& batch, size_t miniBatchSize, size_t numEpochs,
 
 	while (status.trainingSizeRemain)
 	{
-		size_t miniBatchSizeActual = std::min(miniBatchSize, status.trainingSizeRemain);
+		size_t miniBatchSizeActual = std::min(m_params.miniBatchSize, status.trainingSizeRemain);
 		arma::mat miniBatch = batch.rows(batchRowIndex, batchRowIndex+miniBatchSizeActual-1);
 		status.trainingSizeRemain -= miniBatchSizeActual;
 		batchRowIndex += miniBatchSizeActual;
@@ -95,7 +95,7 @@ void Rbm::train(const arma::mat& batch, size_t miniBatchSize, size_t numEpochs,
 		arma::mat hid_state(miniBatchSizeActual, m_w.n_cols);
 		arma::mat hid_probs(miniBatchSizeActual, m_w.n_cols);
 
-		for (epoch=0; epoch < numEpochs; epoch++)
+		for (epoch=0; epoch < m_params.numEpochs; epoch++)
 		{
 
 			// Create hidden layer base on training data

source/Rbm.hpp

@@ -35,6 +35,8 @@ public:
 		, gibbsDoSampleHidden(true)
 		, doSampleBatch(false)
 		, numGibbs(1)
+		, miniBatchSize(100)
+		, numEpochs(1000)
 		{
 		}
 
@@ -50,6 +52,8 @@ public:
 			params["gibbsDoSampleHidden"] = (int)gibbsDoSampleHidden;
 			params["doSampleBatch"] = (int)doSampleBatch;
 			params["numGibbs"] = (int)numGibbs;
+			params["miniBatchSize"] = (int)miniBatchSize;
+			params["numEpochs"] = (int)numEpochs;
 
 			return params;
 		}
@@ -64,6 +68,8 @@ public:
 			gibbsDoSampleHidden = params["gibbsDoSampleHidden"] == 1;
 			doSampleBatch = params["doSampleBatch"] == 1;
 			numGibbs = params["numGibbs"].asUInt();
+			miniBatchSize = params["miniBatchSize"].asUInt();
+			numEpochs = params["numEpochs"].asUInt();
 		}
 		
 		double weightDecay;
@@ -74,6 +80,8 @@ public:
 		bool gibbsDoSampleHidden;
 		bool doSampleBatch;
 		size_t numGibbs;
+		size_t miniBatchSize;
+		size_t numEpochs;
 	};
 
 	struct Status
@@ -114,7 +122,7 @@ public:
 	virtual ~Rbm();
 
 	void weightsInit(double stddev, double mu=0.0);
-	void train(arma::mat const &batch, size_t miniBatchSize, size_t numEpochs, IListener *pListener);
+	void train(arma::mat const &batch, IListener *pListener);
 	
 	arma::mat toHiddenState(const arma::mat &visible) const;
 	arma::mat toVisibleState(const arma::mat &hidden) const;
@@ -127,8 +135,12 @@ public:
 	Json::Value toJson() const;
 	void fromJson(Json::Value params);
 	
+	Params& params()
+	{
+		return m_params;
+	}
 private:
-	const Params m_params;
+	Params m_params;
 	arma::mat sample(arma::mat const &src);
 	static arma::mat probsLogistic(arma::mat const &src);
 	void uniform(arma::mat &srcDst, double stdDev=1.0, double mu=0.5);

source/Stack.cpp

@@ -153,17 +153,17 @@ bool Stack::saveWeights()
 	return true;
 }
 
-void Stack::train(const arma::mat& batch, size_t miniBatchSize, size_t numEpochs, Rbm::IListener* pListener)
+void Stack::train(const arma::mat& batch, Rbm::IListener* pListener)
 {
 	Layer *pLayer = m_pLayers;
 	while(pLayer)
 	{
-		train(pLayer->id(), batch, miniBatchSize, numEpochs, pListener);
+		train(pLayer->id(), batch, pListener);
 		pLayer = pLayer->upper;
 	}
 }
 
-void Stack::train(size_t layerId, const arma::mat& batch, size_t miniBatchSize, size_t numEpochs, Rbm::IListener* pListener)
+void Stack::train(size_t layerId, const arma::mat& batch, Rbm::IListener* pListener)
 {
 	arma::mat thisBatch = batch;
 	Layer *pLayer = m_pLayers;
@@ -180,7 +180,7 @@ void Stack::train(size_t layerId, const arma::mat& batch, size_t miniBatchSize,
 	if (pLayer)
 	{
 		std::cout << m_prjname << ": " << " Training of layer " << std::to_string(layerId) << std::endl;
-		pLayer->train(thisBatch, miniBatchSize, numEpochs, pListener);
+		pLayer->train(thisBatch, pListener);
 	}
 }
 

source/Stack.hpp

@@ -30,8 +30,8 @@ public:
 	void addLayer(Layer *pLayer);
 	Layer* getLayer(size_t layerId) const;
 
-	void train(size_t layerId, const arma::mat& batch, size_t miniBatchSize, size_t numEpochs, Rbm::IListener* pListener);
-	void train(const arma::mat& batch, size_t miniBatchSize, size_t numEpochs, Rbm::IListener* pListener);
+	void train(size_t layerId, const arma::mat& batch, Rbm::IListener* pListener);
+	void train(const arma::mat& batch, Rbm::IListener* pListener);
 	bool load();
 	bool save();
 	void weightsInit(double stddev);

source/main.cpp

@@ -101,6 +101,9 @@ int main()
 		stack.addLayer(layer);
 	}
 
+	Layer *layer = stack.getLayer(0);
+	layer->params().learningRate = 0.02;
+
 	// Save project
 	stack.save();
 
@@ -119,12 +122,11 @@ int main()
 #endif	
 	
 	// Train stack
-	stack.train(batch, 100, 1000, &statusDisplay);
+	stack.train(batch, &statusDisplay);
 	
 	// Save weights
 	stack.saveWeights();
 
-	Layer *layer = stack.getLayer(0);
 	arma::mat v = arma::randu(numTraining, layer->bv().n_elem);
 	arma::mat h = layer->toHiddenProbs(v);
 	arma::mat r = layer->toVisibleProbs(h);

test.prj

@@ -15,8 +15,10 @@
                   "doSampleBatch" : 0,
                   "gibbsDoSampleHidden" : 1,
                   "gibbsDoSampleVisible" : 0,
-                  "learningRate" : 0.10000000000000001,
+                  "learningRate" : 0.02,
+                  "miniBatchSize" : 100,
                   "momentum" : 0.5,
+                  "numEpochs" : 1000,
                   "numGibbs" : 1,
                   "weightDecay" : 0
                }