Optimal Estimation in Sensory Systems

摘要: A variety of experimental studies suggest that sensory systems are capable performing estimation or decision tasks at near-optimal levels. In this chapter, I explore the use optimal in describing computations brain. define what is meant by optimality and provide three quite different methods obtaining an estimator, each based on assumptions about nature information available to constrain problem. then discuss how biological might go computing (and learning compute) estimates. The brain awash signals. How does it interpret these signals, so as extract meaningful consistent environment? Many require esti- mation environmental parameters, there substantial evidence system representing extracting very precise estimates parameters. This particularly impressive when one considers fact built from a large num- ber low-energy unreliable components, whose responses affected many extraneous factors (e.g., temperature, hydration, blood glucose oxygen levels). problem well studied statistics engineering communi- ties, where plethora tools have been developed for designing, implementing, calibrating testing such systems. recent years, used benchmarks models perception. Specifically, development signal detection theory led widespread statistical framework assessing performance perceptual experiments. More recently, (in particular, Bayesian estimation) has human tasks.