Quick ML question for anyone who has more experience than I do. What would be the best way to machine learn a set of features where for each sample that maps to a response class, you have multiple sets of information. To be more clear, think about a molecule from statistical mechanics. It will occupy many different states with different probabilities. So let's say for a sample, aka a molecule, I look at different features of that molecule in those different states. I would for that sample have sets of information. This is unfortunately not like a time series problem. Only thought thus far is to create features that correspond to observables like ensemble average, max, min, stdev etc. Any help would be appreciated!