Figure 5From: Learning without lossTime series of \(\mathtt {RRR\_err}\), the distance between constraint sets A and B, for three runs of the algorithm on the \(m=6\) LEDM instance. As \(\mathtt {RRR\_err}\) is analogous to loss, we see that RRR has no trouble negotiating many local minima in the course of finding solutionsBack to article page