Tutorial on Pattern Classification in Cell Recording

In this chapter, we outline a procedure to decode information from multivariate neural data. We assume that neural recordings have been made from a number of trials in which different conditions were present, and our procedure produces an estimate of how accurately we can predict the labels of these conditions in a new set of data. We call this estimate of future prediction the “decoding/readout accuracy,” and based on this measure we can make inferences about what information is present in the population of neurons and on how this information is coded. The steps we cover to obtain a measure of decoding accuracy include: (1) designing an experiment, (2) formatting the neural data, (3) selecting a classifier to use, (4) applying cross-validation to random splits of the data, (5) evaluating decoding performance through different measures, and (6) testing the integrity of the decoding procedure and significance of the results. We also discuss additional topics, including how to examine questions about neural coding and how to evaluate whether the population is representing stimuli in an invariant/abstract way. Chapter 18 (Singer and Kreiman) discusses statistical classifiers in further detail. The ideas discussed here are applied in several chapters in this book including chapter 2 (Nirenberg), chapter 3 (Poort, Pooresmaeili, and Roelfsema), chapter 7 (Pasupathy and Brincat), chapter 10 (Hung and DiCarlo), chapter 21 (Panzeri and Ince), and chapter 22 (Berens, Logothetis, and Tolias). Chapter 20 discusses related concepts within the domain of functional imaging.