This allows us to determine an optimal batch size for a single experiment, and scale it to determine batch sizes for any number of varied experiments, optimizing test accuracy across the board.ConclusionThis is just a taste of the breadth (and depth) of classification problems in modern machine learning research — everything in this article was published in the last few months..Hopefully, this shows you a little bit of the exciting complexity that one might consider when choosing their classification models in the future!. More details