BIC: Neural Computation That Supports Multiple Cognitive Tasks

BIC：支持多种认知任务的神经计算

• 批准号: 0432037
• 负责人: Leslie Valiant
• 金额: $ 20万
• 依托单位: Harvard University
• 依托单位国家: 美国
• 项目类别: Continuing Grant
• 财政年份: 2004
• 资助国家: 美国
• 起止时间: 2004-08-15 至 2008-07-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=0432037&HistoricalAwards=false
• 关键词: BIC; Neural; Computation; Supports; Multiple

The computational resources available to the brain can be summarized in terms of a number of numerical parameters. These include the number of neurons, the average number of neurons each neuron synapses with, and the basic response time of a neuron to its inputs. A fourth parameter is the strength of influence that a neuron has through its synapses on a typical neuron with which it synapses. Given the limited resources that the brain has, as summarized by these four parameters as currently understood, it turns out to be highly challenging to explain how the brain can perform even the most basic information processing tasks. In particular, the sparsity of the interconnection pattern in conjunction with generally weak synapses appears to impose very considerable constraints. The aim of the proposed research is to explore, using abstract computational models, how information processing is possible at all under these constraints. In particular the challenge is to explain how multiple tasks, such as memorization, association and inductive learning, can be supported together in such a system.

大脑可用的计算资源可以用一些数值参数来概括。这些包括神经元的数量，每个神经元与之突触的平均神经元数量，以及神经元对其输入的基本响应时间。第四个参数是一个神经元通过其突触对与它建立突触的典型神经元的影响强度。鉴于大脑的资源有限，正如我们目前所理解的这四个参数所概括的那样，要解释大脑是如何完成最基本的信息处理任务的，是一项极具挑战性的任务。特别是，连接模式的稀疏性与一般弱突触相结合，似乎施加了相当大的限制。本研究的目的是利用抽象的计算模型来探索在这些约束条件下信息处理是如何可能的。特别的挑战是解释多重任务，如记忆、联想和归纳学习，如何在这样一个系统中一起得到支持。