Parallel Computation and Learning

并行计算和学习

• 批准号: 9200884
• 负责人: Leslie Valiant
• 金额: $ 27万
• 依托单位: Harvard University
• 依托单位国家: 美国
• 项目类别: Continuing Grant
• 财政年份: 1992
• 资助国家: 美国
• 起止时间: 1992-09-15 至 1996-02-29
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=9200884&HistoricalAwards=false
• 关键词: Parallel; Computation; Learning

The research is in two areas: general purpose parallel computing, and models of neural computation. The first area will investigate the possibility of finding a standardizing bridging model for parallel computing, at a level intermediate between the architecture and language levels. Such a standard would insulate the development of software and hardware and thereby speed the success of the parallel computing industry. The bulk-synchronous parallel (BSP) model will be considered as the prime candidate. The second area of research will further develop a discrete model of neural computation that is both weak enough that its constituents are plausibly simulated by human cortex, yet strong enough that some basic tasks of learning and memory can be implemented on it.

该研究涉及两个领域：通用并行计算和神经计算模型。 第一个领域将研究在架构和语言级别之间的中间级别上寻找并行计算标准化桥接模型的可能性。 这样的标准将隔离软件和硬件的开发，从而加速并行计算行业的成功。 批量同步并行（BSP）模型将被视为主要候选模型。 第二个研究领域将进一步开发神经计算的离散模型，该模型既足够弱，其组成部分可以由人类皮层模拟，又足够强大，可以在其上实现一些基本的学习和记忆任务。