fizzbuzz.py

import tensorflow as tf

def encode_input(num):
    return [int(i) for i in tuple(bin(num)[2:].zfill(8))]

def encode_fizzbuzz(num):
    if num % 3 == 0 and num % 5 == 0:
        # Fizzbuzz
        return [1, 0, 0, 0]
    elif num % 3 == 0:
        # Fizz
        return [0, 1, 0, 0]
    elif num % 5 == 0:
        # Buzz
        return [0, 0, 1, 0]
    else:
        # Number
        return [0, 0, 0, 1]

def decode_fizzbuzz(result, num):
    return ['FizzBuzz', 'Fizz', 'Buzz', num][max(range(len(result)), key=lambda x: result[x])]

def main():
    x = []
    y = []

    model = tf.keras.Sequential([
        tf.keras.layers.Dense(64, input_dim=8, activation='relu'),
        tf.keras.layers.Dense(128, activation='relu'),
        tf.keras.layers.Dense(4, activation='softmax')
    ])

    model.compile(tf.keras.optimizers.Adam(learning_rate=0.001),
                  loss='categorical_crossentropy')

    for i in range(100):
        x.append(encode_input(i))
        y.append(encode_fizzbuzz(i))

    print(x,y)

    model.fit(x=x, y=y,
            verbose=2, shuffle=True, epochs=1000)

    correct = 0
    wrong = 0
    for i in range(1, 101):
        result = model.predict([encode_input(i)])
        output = decode_fizzbuzz(result[0], i)
        print(output, end=' ')
        if output == decode_fizzbuzz(encode_fizzbuzz(i), i):
            correct += 1
        else:
            wrong += 1

    print('')
    print('Total correct:', correct)
    print('Wrong:', wrong)
    print('Correct percentage:', correct / (correct + wrong) * 100, '%')


if __name__ == '__main__':
    main()