This study investigates a population decoding paradigm in which the maximum
likelihood inference is based on an unfaithful decoding model (UMLI). This
is usually the case for neural population decoding because the encoding pr
ocess of the brain is not exactly known or because a simplified decoding mo
del is preferred for saving computational cost. We consider an unfaithful d
ecoding model that neglects the pair-wise correlation between neuronal acti
vities and prove that UMLI is asymptotically efficient when the neuronal co
rrelation is uniform or of limited range. The performance of UMLI is compar
ed with that of the maximum likelihood inference based on the faithful mode
l and that of the center-of-mass decoding method. It turns out that UMLI ha
s advantages of decreasing the computational complexity remarkably and main
taining high-level decoding accuracy Moreover, it fan be implemented by a b
iologically feasible recurrent network (Pouget, Zhang, Deneve, & Latham, 19
98). The effect of correlation on the decoding accuracy is also discussed.