def cost_func(x): return softmax.cost(x, num_classes, input_size, lamb, input_data, labels)[0]
input_size = 8
# only 100 datapoints
input_data = input_data[:, :100]
labels = labels[:100]
# only top input_size most-varying input elements (pixels)
indices = input_data.var(1).argsort()[-input_size:]
input_data = np.asfortranarray(input_data[indices, :])
# Randomly initialise theta
theta = 0.005 * np.random.randn(num_classes * input_size, 1)
# === Step 2: Implement softmaxCost ===
#
# Implement softmaxCost in [softmax.cost()](softmax.html#section-1).
cost, grad = softmax.cost(theta, num_classes, input_size, lamb, input_data, labels)
# === Step 3: Gradient checking ===
#
# As with any learning algorithm, always check that the gradients are
# correct before learning the parameters.
# Cost function
def cost_func(x):
return softmax.cost(x, num_classes, input_size, lamb, input_data, labels)[0]
# For testing…
if False:
num_grad = util.compute_numerical_gradient(cost_func, theta)
num_grad = num_grad.ravel('F')