Patch2Self: Supplement ShreyasFadnavis1, JoshuaBatson2, EleftheriosGaryfallidis1 1Indiana UniversityBloomington, 2CZBiohub 1 RegressionModelSelection: SelfSupervisedLoss

From Figure 1,we can see that OLS and Ridge givevery similar performance and the Lasso performs slightly worse. The [9] data set has 160 volumes and the loss was computed for each one of them. For both Ridge and Lasso, the regularization parameteralphawas set to '1.0'. One can see that if the rank is too low (ranks: 1 to 4), a lot of structure is lost and the image has a lot ofinduced smoothness. The raw and denoised outputs from the software mentioned in Sec.