The sign

[Airuibadi]

pytorch_influence_functions/pytorch_influence_functions/calc_influence_function.py

Lines 258 to 271 in 4df5d2e

	tmp_influence = -sum(
	[
	###################################
	# TODO: verify if computation really needs to be done
	# on the CPU or if GPU would work, too
	###################################
	torch.sum(k * j).data.cpu().numpy()
	for k, j in zip(grad_z_vecs[i], e_s_test)
	###################################
	# Originally with [i] because each grad_z contained
	# a list of tensors as long as e_s_test list
	# There is one grad_z per training data sample
	###################################
	]) / train_dataset_size

In the final step， the code here accually calculate the -1/n·I_up,loss(z,z_test). However, in the equation(2) of the original paperhttps://arxiv.org/abs/1703.04730, the term I_up,loss(z,z_test) do has a minus sign. So two negatives make a positive， the calculating here should be 1/n·I_up,loss(z,z_test). Or I just misunderstand some part of code or paper?