Fig. 7: Holistic/quasi-holistic bursting cells carry perfect information of trained complex sounds.
From: Single-neuron representation of learned complex sounds in the auditory cortex
a Cartoon illustration of the machine classification task. Note that in practice, the biological experiments and the machine classifications were individually and sequentially performed, but the logic of information flow was as shown in the model. b Left cartoon: illustration of a scenario of a task with BBN-trained mice. Middle and right graphs: accuracy of classification (Ac) as a function of the neuronal population size (Np) for one example animal. Coloured lines and shades indicate the mean and variance in the Ac value corresponding to each Np value. Inset panels outlined by dashed boxes show the magnified views of the Ac–Np relation curve near the transition point. c Same analysis as in (b) performed in a hypermouse (as illustrated in the left cartoon). A neuronal response pattern of a given trial is made by concatenating the patterns of all involved FOVs of the tested mice together; patterns in each FOV for the same sound with the same chronologically ordered trial index were put together. d Same analysis as in panel c but for a different hypermouse dataset from the chord-trained animals. Shading indicates s.d.
