added label

src/rbm/label.py

@@ -1,10 +1,66 @@
+from rbm.matrix import Mat, np
+from rbm.model import Model
+from rbm.entity import Entity, EntityParams
+from rbm.train import TrainingParams
+
+import math
 class Label:
-	def __init__(self, voc_size=1001, num_codes_per_dim=10, v_min=0.0, v_max=1.0):
+	class Fitter(Model):
-		self.num_codes_per_dim = num_codes_per_dim
+		def __init__(self, dim: tuple[int,int], work_dir: str = '.'):
-		self.voc_size = voc_size
+			super().__init__(f"label-{dim[0]}x{dim[1]}", work_dir)
-		self.v_min = v_min
+			self.unit1 = Entity(dim, EntityParams(num_gibbs_samples=1))
-		self.v_max = v_max
+
-		self.num_dim = Label.num_dims(voc_size, num_codes_per_dim)
+		def forward(self, x: Mat) -> Mat:
+			return self.unit1.forward(x)
+
+		def reconstruct(self, x: Mat):
+			return self.unit1.reconstruct(x)
+
+	def __init__(self, voc_size, num_dim, work_dir):
+		self.num_bits = round(math.log(voc_size,2))
+		self.fitter = Label.Fitter((self.num_bits, num_dim), work_dir)
+
+	def label2vec(self, list_of_labels: np.array):
+		batch = np.zeros((len(list_of_labels), self.num_bits))
+		for index, label in enumerate(list_of_labels):
+			b = self._enc_binary(label, self.num_bits)
+			batch[index, :] = np.array(b)
+
+		return batch
+
+	def vec2label(self, label_vecs: np.array, thresh=0.9):
+		labels = np.zeros((len(label_vecs)))
+		for index, label_vecs in enumerate(label_vecs):
+			labels[index] = self._dec_binary(label_vecs > thresh)
+		return labels
+
+	def fit(self, list_of_labels: np.array):
+		label_vecs = self.label2vec(list_of_labels)
+		self.fitter.train(label_vecs, TrainingParams(num_epochs=10000, do_rao_blackwell=False))
+
+	def encode(self, list_of_labels: np.array):
+		label_vecs = self.label2vec(list_of_labels)
+		enc_data = self.fitter.forward(label_vecs)
+		return enc_data, label_vecs
+
+	def decode(self, list_of_encoded_label: np.array):
+		dec_data = self.fitter.reconstruct(list_of_encoded_label)
+		return dec_data
+
+	@staticmethod
+	def _enc_binary(number: int, num_bits):
+		res = [1 if s == '1' else 0 for s in format(number, f'{num_bits}b')]
+		return res
+
+	@staticmethod
+	def _dec_binary(vec: np.array):
+		res = 0
+		e = 1
+		for v in reversed(vec):
+			res += v*e
+			e *= 2
+
+		return res
 
 	@staticmethod
 	def num_dims(voc_size, num_codes_per_dim):
@@ -17,5 +73,33 @@ class Label:
 
 
 if __name__ == "__main__":
-	enc = Label()
+	voc_size = 100
+
+	# Create Labeler
+	enc = Label(voc_size, 16, work_dir="../../results")
+
+	# Load
+	enc.fitter.load()
+
+	# Train
+	train_labels = list(range(voc_size))
+	enc.fit(train_labels)
+
+	# Save fitter state
+	enc.fitter.save()
+
+	# Encode test labels
+	test_labels = np.array([1, 23, 99, 37, 55, 7, 31, 10, 19, 70])
+	encoded, binary_labels = enc.encode(train_labels)
+
+	# Decode
+	decoded_soft = enc.decode(encoded)
+	decode_hard = (decoded_soft > 0.9).astype(int)
+
+	# Convert soft state into labels
+	labels_reconst = enc.vec2label(decoded_soft)
+	print(labels_reconst)
+
+	print(f"Soft Error: {np.sum((decoded_soft - binary_labels) ** 2)}")
+	print(f"Hard Error: {np.sum((decode_hard - binary_labels) ** 2)}")
 	print("Test: [passed]")

src/tests/test_label.py

@@ -0,0 +1,45 @@
+import matplotlib.pyplot as plt
+
+from rbm.matrix import np
+from rbm.label import Label
+
+if __name__ == "__main__":
+	voc_size = 100
+	do_train = False
+	work_dir = "../../results"
+	prj_root = "/home/jens/work/repos/Rbm"
+
+	# Create Labeler
+	enc = Label(voc_size, 16, work_dir)
+
+	# Load
+	enc.fitter.load()
+
+	# Train
+	if do_train:
+		train_labels = list(range(voc_size))
+		enc.fit(train_labels)
+		enc.fitter.save()
+
+	# Encode test labels
+	test_labels = np.array([1, 23, 99, 37, 55, 7, 31, 10, 19, 70])
+	encoded, binary_labels = enc.encode(test_labels)
+
+	# Decode
+	decoded_soft = enc.decode(encoded)
+	decode_hard = (decoded_soft > 0.90).astype(int)
+
+	print(f"Soft Error: {np.sum((decoded_soft - binary_labels) ** 2)}")
+	print(f"Hard Error: {np.sum((decode_hard - binary_labels) ** 2)}")
+
+	fig, axes = plt.subplots(1, len(encoded), figsize=(12, 3))
+	for index, inp in enumerate(encoded):
+		img = np.reshape(inp, (4, 4))
+		axes[index].imshow(img)
+		axes[index].axis('off')
+		axes[index].set_title(f'{test_labels[index]}')
+
+	plt.show()
+
+	print("Test: [passed]")
+

src/tests/test_norbs.py

@@ -1,5 +1,4 @@
 import os
-
 import matplotlib.pyplot as plt
 
 from rbm.model import Model
@@ -7,8 +6,6 @@ from rbm.entity import Entity, EntityParams
 from rbm.matrix import Mat, np, read_armadillo
 from rbm.train import TrainingParams
 
-import matplotlib.pyplot as plot
-
 class TestModel(Model):
 	def __init__(self, name: str, work_dir: str = '.'):
 		super().__init__(name, work_dir)