The notebook sometimes recommends incorrect/bad dimensionality. In some cases, it will recommend high dimensionality even when the number of robust components drop sharply (see image). In these high dimensional decomposition (almost) all iModulons have near 0 activities in all samples.
The notebook sometimes recommends incorrect/bad dimensionality. In some cases, it will recommend high dimensionality even when the number of robust components drop sharply (see image). In these high dimensional decomposition (almost) all iModulons have near 0 activities in all samples.