A Mixed Filter Algorithm for State Estimation from Simultaneously Recorded Continuous-Valued Point Process and Binary Observations

Learning is a dynamic process generally defined as a change in behavior as a result of experience. Understanding how processes at the molecular and neuronal levels integrate so that an organism can learn is a central question in neuroscience. Most learning experiments consist of a sequence of trials. During each trial, a subject is given a fixed amount of time to execute a task and the resulting performance is recorded. During each trial, performance can be measured with not only binary modalities corresponding to whether or not the subject executed a task correctly, but also continuous-valued measurements - such as the reaction time, the time it takes the subject to respond, or the run the time, the total amount of time it takes the subject to execute a task. Neurophysiological measures that are recorded - such as the spiking behavior of certain neurons - can also characterize learning. Learning is usually illustrated by showing that the subject has successfully performed the previously unfamiliar chance with greater reliability than simply by chance.