Optimization of decision making in multilayer networks: The role of locus coeruleus

摘要: Previous theoretical work has shown that a single-layer neural network can implement the optimal decision process for simple, two-alternative forced-choice (2AFC) tasks. However, it is likely mammalian brain comprises multilayer networks, raising question of whether and how performance be approximated in such an architecture. Here, we present suggesting noradrenergic nucleus locus coeruleus (LC) may help optimize 2AFC making brain. This based on observations neurons LC selectively fire following presentation salient stimuli tasks corresponding release norepinephrine transiently increase responsivity, or gain, cortical processing units. We describe computational simulations investigate role gain changes optimizing making. In model, no prior cueing knowledge stimulus onset time assumed. Performance assessed terms rate correct responses over (the reward rate). first results model accumulates (integrates) sensory input implements as threshold crossing. Gain transients, representing modulatory effect LC, are driven by separate crossings this layer. all free parameters to determine maximum achievable compare when held fixed. find dynamic mechanism yields improvement model. We then examine two-layer which competing accumulators layer (capable implementing task relevant decision) pass activity response second Again, version crossing (decision) elicits (and concomitant gain) with fixed-gain model. transients modeling phasic yield 12% 24%. Furthermore, show timing characteristics these agree concerning firing patterns reported recent experimental studies. provides converging evidence hypothesis optimizes processes underlying networks.