Keras Deep Learning Cookbook

Encoder

1. First, we define an encoder, which converts char to a one-hot encoded value in an array of length 91:

def encode(pattern, n_unique):    encoded = list()    for value in pattern:        row = [0.0 for x in range(n_unique)]        index = ord(value)        row[ord(value)] = 1.0        encoded.append(row)    return encoded

2. Divide the sequence into X and y values as follows:

def to_xy_pairs(encoded):    X,y = list(),list()    for i in range(1, len(encoded)):        X.append(encoded[i-1])        y.append(encoded[i])    return X, y

3. Convert the x, y values into a three-dimensional matrix which LSTM can understand, as follows: 

def to_lstm_dataset(sequence, n_unique):    # one hot encode    encoded = encode(sequence, n_unique)    # convert to in/out patterns    X,y = to_xy_pairs(encoded)    # convert to LSTM ...