- added gibbs sampling for forward pass

src/rbm/layer.py

@@ -13,6 +13,9 @@ class RbmLayer:
 		self.params = params
 		self.state_filename = f"{self.name}_state.npz"
 
+	def init(self):
+		self.state.init()
+
 	def save(self):
 		self.state.to_file(self.state_filename)
 
@@ -89,6 +92,22 @@ class RbmLayer:
 		return prob(state)
 
 
+	def gibbs_v_to_h(self, v: np.ndarray) -> np.ndarray:
+		h = None
+		for i in range(self.params.num_gibbs_samples):
+			h = self.v_to_ph(v)
+			v = self.h_to_pv(h)
+
+		return h
+
+	def gibbs_h_to_v(self, h: np.ndarray) -> np.ndarray:
+		v = None
+		for i in range(self.params.num_gibbs_samples):
+			v = self.h_to_pv(h)
+			h = self.v_to_ph(v)
+
+		return v
+
 def xor():
 	params = RbmParams()
 	params.do_rao_blackwell = True
@@ -96,15 +115,24 @@ def xor():
 	params.mini_batch_size = 100
 	params.learning_rate = 0.1
 	params.momentum = 0.5
-	params.num_epochs = 10000
+	params.num_epochs = 100
 
 	status = Status()
 	layer = RbmLayer("Layer_0", 3, 16, params)
-	training_batch = np.array([[0,0,0], [0,1,1], [1,0,1], [1,1,0]], dtype=np.float64)
+	layer.init()
-	test_batch = np.array([[0,0,0], [0,1,0], [1,0,0], [1,1,0]], dtype=np.float64)
 
+	# Train
+	training_batch = np.array([[0,0,0], [0,1,1], [1,0,1], [1,1,0]], dtype=np.float64)
 	layer.train(training_batch, cd_jens, status)
 
+	# Test
+	test_batch = np.array([[0,0,0], [0,1,0], [1,0,0], [1,1,0]], dtype=np.float64)
+	for pattern in test_batch:
+		h = layer.gibbs_v_to_h(pattern)
+		v = layer.gibbs_h_to_v(h)
+		print(f"P{pattern} : {v}")
+
+
 if __name__ == "__main__":
 	xor()