Active Measurement for Neuroscience

Ross Boczar, UC Berkeley

Abstract:  In neuroscience we often wish to measure the properties of individual neurons or groups of neurons in  response to certain stimuli. These experiments are expensive and time-limited, as all biological samples decay and ultimately die. Thus, for a given neuron we would often like to know the best stimulus to present to learn a certain property. In this preliminary work, we explore the possibility of exploiting elastic compute resources for active measurement selection.