All the listed loss functions induce equivalent topologies. Functions which converge in absolute value on a compact space converge also with mean squared error to the same function. The loss function is actually irrelevant and all it does is change the rate of convergence.