Hello !
I just finished “Advanced Leaning Algorithms” on coursera and would like to put my knowledge into practice.
To do this, I would like to make an algorithm that can predict that for example:
The sequence of [42, 43, 44] is [45, 46, 47].
I thought it would be very simple.
x = np.array([
[1, 2, 3],
[2, 3, 4],
[3, 4, 5],
[4, 5, 6],
[5, 6, 7],
[6, 7, 8],
[7, 8, 9],
[8, 9, 10],
[9, 10, 11],
[10001, 10002, 10003]
])
y = np.array([
[4, 5, 6],
[5, 6, 7],
[6, 7, 8],
[7, 8, 9],
[8, 9, 10],
[9, 10, 11],
[10, 11, 12],
[11, 12, 13],
[12, 13, 14],
[10004, 10005, 10006]
])
to_predict = [[42, 43, 44]]
I have tried several models that are very, very far from being right, while I thought that this simple model would allow to have zero loss.
model = tf.keras.models.Sequential([
tf.keras.layers.Dense(2, activation='linear', input_shape=(3,)),
tf.keras.layers.Dense(3, activation='linear')
])
model.compile(optimizer='adam', loss='mse')
model.fit(x, y, epochs=478)
result = model.predict(np.array(to_predict))
The example above gives me a more or less approximate result, but it’s fair if I put thousands of examples during the training.
It really amazes me, given the simplicity of the statement, to have a model that can’t get it right, and to have so many examples needed.
Do you have any advice for me?
Thank you very much !