BioComp: Translating Mechanisms of Novelty Recognition in Drosophila into a Computational Device

点击翻译按钮获取中文摘要

Fruit flies exhibit versatile and sophisticated capabilities of stimulus discrimination and"attention"-like behavior. They are particularly attuned to recognizing novelty. This proposaloutlines a plan to delineate the network and neural principles underlying the fruit fly's ability toperform these information-processing functions, and then to employ them as the basis for acomputational device. This multi-step program will be begun by defining the circuitry in the fruitfly subserving its recognition, selective attention, and novelty responses, analyzing thecontributions, connections, and interactions of these circuit elements both behaviorally andphysiogically, and then introducing this neural architecture as the basis for the computer-simulatedbrain in a brain-based device capable of displaying novelty recognition.We will use techniques of gene targeting that we have developed previously to identifythe parts of the brain contributing to recognition, selective attention, and novelty responses. Wewill map the sites in the nervous system mediating the effect by manipulating neural activity.This will be achieved in two opposing ways: one way by blocking activity and the other byincreasing activity. These perturbations will be targeted to different, restricted parts of the brainby means of a set of genetically engineered fly strains we have developed and used over theyears. In this manner, we will map the funcitonal circuitry mediating a fruit fly's noveltyresponse by increasing or decreasing excitability in restricted brain regions.Aim 1: Analyze the complex, organizational architecture by which the fruit fly's nervous systemachieves behaviorally the recognition of novelty.Aim 2: Map the distribution of the 20-30 Hz LFP response in various brain regions, and the roleof coherence between these brain regions in the generation of the fruit fly's novelty response.Brain-based devices provide the groundwork for the development of intelligent machinesthat follow neurobiological rather than computational principles in their construction. As is thecase with animals, the behaviors of brain-based devices emerge solely as a result of internallygenerated activity of their nervous systems rather than of responses to any programmedinstructions from computer software. Such devices are particularly useful in situations of noveltywhere computation is not possible in principle or in cases of great local complexity whereprogramming proves infeasible. Such a device must confront novel situations and complex sets ofparameters that must be dealt with rapidly. Our goal is to implement principles from the fruit flysystem into such a device, using as our platform an existing brain-based device developed at TheNeurosciences Institute.Aim 3: Introduce a simulated neural architecture based on the functional network for noveltydetection defined in Aims 1 & 2 into a brain-based device.The long-term goal of this work is to understand the principles upon which the nervoussystem of the fruit fly operates as the basis for neurobiologically inspired computational devices.The principles underlying nervous system function hold promise for developing a new generationof devices that would be more capable of adaptive behavior than current systems. The mostsophisticated behavior seen in either biological or artificial agents is shown by organisms whosebehavior is guided by a nervous system. The fruit fly offers the requisite complexity to be ofvalue in this endeavor, while being simple enough (i.e., small enough in neuron number) to beamenable to analysis. Most importantly, it offers sophistication afforded by a multi-disciplinaryexperimental approach (genetics, anatomy, physiology and behavior) to be followed by computersimulation and device implementation.