Neural Population Decoding in Short-Time Windows

External information is encoded in spiking activities of neural population. The present study investigates the performance of population decoding in a short-time window. Two decoding strategies, namely, maximum likelihood inference and template-matching, are explored. We find that in a short-time window, two methods are not efficient and that their errors satisfy the Cauchy distributions. As expected, maximum likelihood inference outperforms template-matching asymptotically. However, in a very short time window, template-matching has smaller decoding errors than maximum likelihood inference. The implication of this result is discussed.