Impact of the reading time window duration on different decoding mechanisms.

<p>(<b>A</b>) Correct stimulus prediction curves during the first breathing cycle (during odorant application) computed for different coding schemes and different time window durations. <i>Black curves</i>: instantaneous firing rate (i.e. in independent time bins). <i>Red curves</i>: concatenated firing over consecutive time bins (temporal sequence analysis). <i>Blue curves</i>: cumulative spike counts over consecutive time bins. Each graph shows the prediction curves computed with a population vector built with different time bin durations (2, 12, 50 and 100 ms). (<b>B</b>) Maximum of prediction in the first breathing cycle computed for different bin durations and for the different coding schemes during odorant presentations. Noisy fluctuations of the prediction curves increase for analysis using short bin duration, leading to an overestimation of the true maximum of prediction. For comparison, the maximum of correct prediction computed on a respiratory cycle before odorant application is plotted. (<b>C</b>) Ratio of the prediction curves for the concatenated code over the cumulative code. For each window duration, a ratio between the prediction curves of the concatenated and cumulative schemes is computed either on the first 100 ms (<i>magenta</i>), 200 ms (<i>green</i>) after odor onset (i.e. for 0 to 100 or 200 respectively) or the entire respiratory cycle duration (<i>cyan</i>). Cumulative code is more informative for shorter time bin duration (<4–6 ms) while concatenated code becomes more informative for longer duration (from 10 to 50 ms).