Continual Learning by Asymmetric Loss Approximation With Single-Side Overestimation | IEEE Conference Publication | IEEE Xplore