Table 1: Synthetic data: Final MISA loss values (lower is better).

Each row represents the ground-truth (GT) subspace structure used to generate the data and each column represents the test subspace structure used to fit the model. The lowest loss value along the row is highlighted in bold, which determines the selected subspace. Approaches performing consistently well in relation to the ISI in Figure 3 will contain bold loss values only along the diagonal. The loss value is largely consistent with the ISI value, except that it incorrectly implies that MSIVA $S_5$ is a better fit when $S_4$ is used to generate the data. Further, the multimodal baseline results incorrectly imply that $S_3$ and $S_1$ are better when $S_1$ and $S_4$ are the ground-truth subspace structures, respectively. Overall, the differences in diagonal loss values between MSIVA and the unimodal baseline appear negligible considering the correspondingly negligible differences in ISI (Figure 3).

Unimodal Baseline	$S_1^{\text{Test}}$	$S_2^{\text{Test}}$	$S_3^{\text{Test}}$	$S_4^{\text{Test}}$	$S_5^{\text{Test}}$
$S_1^{\text{GT}}$	42.692	42.884	42.762	42.992	43.230
$S_2^{\text{GT}}$	42.649	42.300	42.851	42.868	42.918
$S_3^{\text{GT}}$	42.720	42.858	42.635	43.100	43.256
$S_4^{\text{GT}}$	43.091	43.239	43.174	42.976	43.507
$S_5^{\text{GT}}$	43.401	43.010	43.497	43.773	42.021
MSIVA	$S_1^{\text{Test}}$	$S_2^{\text{Test}}$	$S_3^{\text{Test}}$	$S_4^{\text{Test}}$	$S_5^{\text{Test}}$
$S_1^{\text{GT}}$	42.677	42.865	42.751	43.038	43.111
$S_2^{\text{GT}}$	42.656	42.229	42.628	42.764	42.749
$S_3^{\text{GT}}$	42.695	42.862	42.620	43.040	43.126
$S_4^{\text{GT}}$	42.689	42.397	41.120	39.937	33.609
$S_5^{\text{GT}}$	43.405	42.966	43.388	43.975	42.005
Multimodal Baseline	$S_1^{\text{Test}}$	$S_2^{\text{Test}}$	$S_3^{\text{Test}}$	$S_4^{\text{Test}}$	$S_5^{\text{Test}}$
$S_1^{\text{GT}}$	23.824	23.947	23.819	24.028	24.274
$S_2^{\text{GT}}$	27.766	27.442	27.803	28.162	28.182
$S_3^{\text{GT}}$	23.931	24.029	23.779	24.036	24.229
$S_4^{\text{GT}}$	17.265	18.660	17.290	17.564	19.731
$S_5^{\text{GT}}$	36.764	36.359	36.758	37.265	35.262