Denoise neural signals using tensor decompositions.
Data should be formatted as an array Y
that is size (N,T,C,R)
where N
is the number of neurons, T
is the number of time points, C
is the number of conditions, and R
is the total number of trials. The fourth index should be padded with NaNs
in cases where each neuron-condition pair did not see the same number of trials. Further, trial ordering conveys no meaning.
tensorDenoiseGridSearchCV.m
-- Finds the optimal tensor rank via grid search and leave-one-out cross-validation.tensorDenoiseCluster.m
-- A wrapper file to pass parameters through tensorDenoiseGridSearchCV.m.tensorDenoiseERR.m
-- Computes the relative error of tensor Yhat relative to Y.tensorDenoiseDat2Plot.m
-- Takes the output files fromtensorDenoiseCluster.m
and produces the cell arrayerrPlot
, which can be used to plot results.