A physiologically-inspired model of numerical classification based on graded stimulus coding.

摘要: In most natural decision contexts, the process of selecting among competing actions takes place in presence informative, but potentially ambiguous, stimuli. Decisions about magnitudes – quantities like time, length, and brightness that are linearly ordered constitute an important subclass such decisions. It has long been known perceptual judgments obey Weber's Law, wherein just-noticeable difference a magnitude is proportional to itself. Current physiologically inspired models numerical classification assume discriminations made via labeled line code neurons selectively tuned for numerosity, pattern observed firing rates ventral intraparietal area (VIP) macaque. By contrast, contiguous lateral (LIP) signal numerosity graded fashion, suggesting possibility could be achieved absence number. Here, we consider performance model based on this analog coding scheme paradigmatic discrimination task bisection. We demonstrate basic two-neuron classifier model, derived from experimentally measured monotonic responses LIP neurons, sufficient reproduce bisection behavior monkeys, threshold can set by reward maximization simple learning rule. addition, our predicts deviations Weber Law scaling choice at high numerosity. Together, these results suggest both generic neuronal framework magnitude-based decisions role contingency