Or you can actually visualize and quantify impact of aggregates of meaningful features within networks using the methodology described in NetDissect or in TCAV.

Or you can actually visualize and quantify impact of aggregates of meaningful features within networks using the methodology described in NetDissect or in TCAV. But of course, that casts doubt into a lot of claimed mechanisms and tons of ML research as I have found out.