Figure 8From: Learning without lossDistribution of final reconstruction errors in 100 runs of RRR and coordinate descent (CD) on the letter-montage data setBack to article page