I have some data on a population of single neurons, and I want to see if the activity pattern observed during a pre-stimulus baseline has a similar structure to that observed when the stimulus is presented. Basically, when the expected stimulus (at baseline) is the same as the stimulus that is shown I expect decoding accuracy to be only modestly lower, but when expected and actual stimulus don’t match I expect a different activity pattern between them, which would translate to decoding performance which is worse than when expectation and stimulus match (i.e. the former condition).

I know how the k-nearest neighbor classifier works, and I know how to implement it. What I don’t know is whether this is actually a common/sensible approach.

I tried to look for publications using such a method but I haven’t found any and frankly I’m not sure whether I’m using the right search terms…

Any help – in the form of explanations, links, literature references, search terms – is greatly appreciated!