Mistake in Shubert function

[gbelouze]

The implementation is not correct. It reads

d = X.shape[0]
for i in range(0,d):
    res = np.prod(np.sum([i * np.cos((j+1)*X[i] + j) for j in range(1, 5+1)]))
return res

I'm guessing the intention was

d = X.shape[0]
res = np.prod([
    np.sum([i * np.cos((j+1)*X[i] + j) for j in range(1, 5+1)]))
    for i in range(d)
])
return res

I suggest the following implementation which avoids loops (inefficient in numpy), and is IMO more readable (it does allocate a bigger array)

def f(self, X):
    j = np.arange(1, 6)[None, :]
    res = np.cos((j + 1) * X[:, None] + j) \
        .sum(axis=1) \
        .prod()
    return res