84 Quantifying and Learning Linear Symmetry-Based Disentanglement (LSBD) methods on the BETA, FACTOR and MOD metrics, and achieves much better scores for the DCI metric where traditional methods scored poorly. 0.00 0.25 0.50 0.75 1.00 1.25 1.50 1.75 2.00 DLSBD 0.0 0.2 0.4 0.6 0.8 1.0 metric value metric Beta Factor SAP DCI MIG MOD (a) Scatter plot. 0.25 0.50 0.75 1.00 1.25 1.50 1.75 DLSBD bin 0.0 0.2 0.4 0.6 0.8 1.0 metric value metric Beta Factor SAP DCI MIG MOD (b) Binned by DLSBDvalue. Figure 4.9: Comparing DLSBD to previous disentanglement metrics. The MIG and SAP scores are still low for methods focusing on LSBD. This is expected however, as explained earlier in Section 4.7.1. This was also observed by Painter et al. (2020) for different datasets.
RkJQdWJsaXNoZXIy MjY0ODMw