I am trying to learn CNN by following stanford’s cs231n lectures and I have a question in assignment 1 of two layer network.
In the affine Affine layer: backward section of assignment1/two_layer_net.ipynb why are we multiplying dout elementwise and summing the differences.
np.random.seed(231)
x = np.random.randn(10, 2, 3)
w = np.random.randn(6, 5)
b = np.random.randn(5)
dout = np.random.randn(10, 5)
dx_num = eval_numerical_gradient_array(lambda x: affine_forward(x, w, b)[0], x, dout)
def eval_numerical_gradient_array(f, x, df, h=1e-5):
"""
Evaluate a numeric gradient for a function that accepts a numpy
array and returns a numpy array.
"""
grad = np.zeros_like(x)
it = np.nditer(x, flags=['multi_index'], op_flags=['readwrite'])
while not it.finished:
ix = it.multi_index
oldval = x[ix]
x[ix] = oldval + h
pos = f(x).copy()
x[ix] = oldval - h
neg = f(x).copy()
x[ix] = oldval
grad[ix] = np.sum((pos - neg) * df) / (2 * h)
it.iternext()
return grad
There is one more function eval_numerical_gradient
calculates loss gradient w.r.t X or W.
oldval = x[ix]
x[ix] = oldval + h # increment by h
fxph = f(x) # evalute f(x + h)
x[ix] = oldval - h
fxmh = f(x) # evaluate f(x - h)
x[ix] = oldval # restore
# compute the partial derivative with centered formula
grad[ix] = (fxph - fxmh) / (2 * h) # the slope
So for each element wise change in X, we get one scalar value of loss so we get dL/dX as matrix of size X, which is understandable.