In the source code, the author calculates the cosine distance as follows.
sum_support = torch.sum(torch.pow(support_image, 2), 1)
support_manitude = sum_support.clamp(eps, float("inf")).rsqrt()
dot_product = input_image.unsqueeze(1).bmm(support_image.unsqueeze(2)).squeeze()
cosine_similarity = dot_product * support_manitude * input_manitude
similarities.append(cosine_similarity)
But in my opinion, the right the cosine distance should be calculated as follows.
sum_support = torch.sum(torch.pow(support_image, 2), 1)
support_manitude = sum_support.clamp(eps, float("inf")).rsqrt()
sum_input = torch.sum(torch.pow(input_image, 2), 1)
input_manitude = sum_input.clamp(eps, float("inf")).rsqrt()
dot_product = input_image.unsqueeze(1).bmm(support_image.unsqueeze(2)).squeeze()
cosine_similarity = dot_product * support_manitude * input_manitude
similarities.append(cosine_similarity)
Am i right? If not, what is the mistake?
In the source code, the author calculates the cosine distance as follows.
But in my opinion, the right the cosine distance should be calculated as follows.
Am i right? If not, what is the mistake?