Tensorflow NN 输入维度

Question

I have an array with a shape of (8, 8, 2) and would like to train a NN on the array. I would like to add one other integer to the training data like so: [array, int] but then end up with a shape like (2,) which I am unable to train with. Is there a way that I can create an array that TensorFlow would accept or train the NN with this new array?

I have tried just making the integer into a vector and adding it onto the array which gives a shape of (9, 8, 2), but that results in a lot of redundant information.

my_array = [[(rook, b), (horse, b), (bishop, b), (queen, b), (king, b), (bishop, b), (horse, b), (rook, b)],
                       [(pawn, b) for i in range(8)],
                       [(0, 0) for i in range(8)],
                       [(0, 0) for i in range(8)],
                       [(0, 0) for i in range(8)],
                       [(0, 0) for i in range(8)],
                       [(pawn, w) for i in range(8)],
                       [(rook, w), (horse, w), (bishop, w), (queen, w), (king, w), (bishop, w), (horse, w), (rook, w)]]


my_int = 1

I_tried = my_array+[[(my_int, my_int) for i in range(8)]]

print(numpy.array(my_array).shape) # --> (8, 8, 2)
print(numpy.array(I_tried).shape) # --> (2,)

Answer 1

You may want to use tuples instead of lists then. Depending on the purpose of this concatenation. But you can try this:

my_int = 1
second_array = [(my_int, my_int) for i in range(8)]

I_tried = (my_array, second_array)