J_1 = 0.5 * torch.norm(output.mT - output.mT@weight, p = ‘fro’)**2
J_2 = torch.sum(torch.sum(abs(weight)))
J_3 = 0.25 * torch.trace((torch.square(output@output.mT - torch.eye(len(output)))))
loss = gamma *J_2 + lamda * J_3+ J_1
The loss function involves two parameters. Since the values of J_1, J_2 and J_3 all have different scale in terms of the magnitude, so the right values of the parameters are very important for loss function to converges. Now without knowing the values of J_1, J_2, and J_3 how to determines the values of the parameters? Is there any general techniques to determine these parameters?