38 rnn

38-rnn/tf-38.py

@@ -0,0 +1,95 @@
+from keras.models import Sequential
+from keras.layers import Dense, LSTM, RepeatVector, TimeDistributed
+from keras.losses import binary_crossentropy, categorical_crossentropy
+from keras.optimizers import SGD
+from keras. metrics import top_k_categorical_accuracy
+from keras import backend as K
+import numpy as np
+import sys, os, string, random
+
+characters = string.printable
+char_indices = dict((c, i) for i, c in enumerate(characters))
+indices_char = dict((i, c) for i, c in enumerate(characters))
+
+INPUT_VOCAB_SIZE = len(characters)
+BATCH_SIZE = 200
+HIDDEN_SIZE = 128
+N_LAYERS = 1
+TIME_STEPS = 3
+
+def encode_one_hot(line):
+    remain = len(line) % TIME_STEPS
+    if remain != 0:
+        line = line + ' ' * (TIME_STEPS-remain)
+    x = np.zeros((len(line), INPUT_VOCAB_SIZE))
+    for i, c in enumerate(line):
+        index = char_indices[c] if c in characters else char_indices[' ']
+        x[i][index] = 1 
+    return np.reshape(x, ( (int)(len(line) / TIME_STEPS), TIME_STEPS, INPUT_VOCAB_SIZE ) )
+
+def decode_one_hot(y):
+    x = np.reshape(y, (y.shape[0]*y.shape[1], INPUT_VOCAB_SIZE))
+    s = []
+    for onehot in x:
+        one_index = np.argmax(onehot) 
+        s.append(indices_char[one_index]) 
+    return ''.join(s)
+    
+def input_generator(nsamples):
+    def generate_line():
+        inline = []; outline = []
+        for _ in range(nsamples):
+            c = random.choice(characters) 
+            expected = c.lower() if c in string.ascii_letters else ' ' 
+            inline.append(c); outline.append(expected)
+        for i in range(nsamples):
+            if outline[i] == ' ': continue
+            if i > 0 and i < nsamples-1:
+                if outline[i-1] == ' ' and outline[i+1] == ' ':
+                    outline[i] = ' '
+            if (i == 0 and outline[1] == ' ') or (i == nsamples-1 and outline[nsamples-2] == ' '):
+                outline[i] = ' '
+        return ''.join(inline), ''.join(outline)
+
+    while True:
+        input_data, expected = generate_line()
+        #print("Input :", input_data)
+        #print("Output:", expected)
+        data_in = encode_one_hot(input_data)
+        data_out = encode_one_hot(expected)
+        yield data_in, data_out
+
+def train(model):
+    model.compile(loss='categorical_crossentropy',
+                  optimizer='adam',
+                  metrics=['accuracy'])
+    input_gen = input_generator(BATCH_SIZE)
+    validation_gen = input_generator(BATCH_SIZE)
+    model.fit_generator(input_gen,
+                epochs = 200, workers=1,
+                steps_per_epoch = 50,
+                validation_data = validation_gen,
+                validation_steps = 10)
+
+def build_model():
+    model = Sequential()
+    r_layer = LSTM(HIDDEN_SIZE, input_shape=(None, INPUT_VOCAB_SIZE))
+    model.add(r_layer)
+    model.add(RepeatVector(TIME_STEPS))
+    for _ in range(N_LAYERS):
+        model.add(LSTM(HIDDEN_SIZE, return_sequences=True))
+    model.add(TimeDistributed(Dense(INPUT_VOCAB_SIZE, activation='softmax')))
+    return model
+
+model = build_model()
+model.summary()
+train(model)
+
+input("Network has been trained. Press <Enter> to run program.")
+with open(sys.argv[1]) as f:
+    for line in f:
+        if line.isspace(): continue
+        batch = encode_one_hot(line)
+        preds = model.predict(batch)
+        normal = decode_one_hot(preds)
+        print(normal)