How to properly transpose matrices in numpy?

Question

Python

Good evening.

For educational purposes, I'm trying to write and train a neural network. And I get the following error:

ValueError: shapes (900,) and (4,) not aligned: 900 (dim 0) != 4 (dim 0)

I understand that it is connected with the dimensions of the matrices, namely, with not respecting their sizes. As far as I understand, numpy's transpose operation does not allow you to convert a row vector to a column vector.

How can I fix this issue?

# количество проходов по учебному набору
iteration = 10

# коэфф. обучения
alpha = 0.004

# инициация размера массива
num_pixels = 900

# инициация размера выхода
num_output = 4

# объявление весов
weight_0_1 = 0.2*np.random.random((num_pixels, num_output)) - 0.1

for j in range(iteration):
    error = 0
    for i in range(len(data)):
        # получение входных значений для первого слоя
        layer_0 = np.array(data[i][1:])
        layer_0 = layer_0.astype(float)
        
        # прямое распространение 
        layer_1 = np.dot(layer_0, weight_0_1)
        
        # ошибка
        error += np.sum((layer_1 - goal_set[i]) ** 2)
        
        # обратное распространение
        layer_1_delta = (goal_set[i] - layer_1)
        weight_0_1 += alpha * np.dot(layer_0.T, layer_1_delta) 
    print(f'I:{j} Error:{error}')

Answer 1

As far as I understand, numpy's transpose operation does not allow you to convert a row vector to a column vector. --- ????? Where did you get this from?

a=np.array([1,2,3,4,5,6,7,8,9,10,11,12,13,14,15])
print(a)
b=a.reshape(-1,1)
print(b)

Result of work:

[ 1  2  3  4  5  6  7  8  9 10 11 12 13 14 15]