An Architecture for Performing Complex Evaluations

用于执行复杂评估的架构

• 批准号: 8709741
• 负责人: Carl Ebeling
• 金额: $ 6.5万
• 依托单位: University of Washington
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 1987
• 资助国家: 美国
• 起止时间: 1987-09-15 至 1990-02-28
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=8709741&HistoricalAwards=false
• 关键词: Architecture; Performing; Complex; Evaluations

Many difficult problems require the analysis of a large number of subproblems and the subsequent combining of the results to form an overall solution. The quality of the solution depends both on how well and how fast this analysis is done. A fast and effective method for performing such analyses in the domain of chess has been described and a prototype implemented in the Hitech chess machine. Hitech, which has surpassed all previous efforts in computer chess by more than 150 rating points, has demonstrated the power of this method in a limited problem domain. However, there remains the open question about the applicability of this evaluation method to other problem areas. The focus of this research is to better understand the power and generality of this architecture with respect to other problems in Artificial Intelligence and on its development of the tools with which new instances of this hardware can be built for other problems. The first goal is being pursued by performing case studies in a number of promising problem areas and proposing solutions based on this architecture. At the same time a general evaluation chip is being designed and implemented that will be targeted for Hitech, but will include the features common to any evaluation chip. The second phase of the research is taking a solution proposed in one of the case studies and constructing a prototype using the tools developed during the design of the first evaluation chip. The project focuses on generalizing a problem solution approach found effective in one problem area to a broader class of problems and evaluating its effectiveness in these new areas. The approach combines this powerful solution approach with modern VLSI technology to produce a unique problem solution capability.

许多难题需要对大量子问题进行分析，然后将结果组合起来形成整体解决方案。 解决方案的质量取决于分析的完成程度和速度。 描述了一种在国际象棋领域执行此类分析的快速有效的方法，并在 Hitech 国际象棋机中实现了原型。 Hitech 已经超越了计算机国际象棋之前所有的努力超过 150 个评级点，在有限的问题领域展示了这种方法的威力。 然而，这种评估方法是否适用于其他问题领域仍然存在悬而未决的问题。 这项研究的重点是更好地理解该架构对于人工智能中其他问题的强大功能和通用性，以及它开发的工具，可以使用这些工具为其他问题构建该硬件的新实例。 第一个目标是通过在许多有前途的问题领域进行案例研究并基于此架构提出解决方案来实现。 同时，正在设计和实现一款通用评估芯片，该芯片将针对 Hitech，但将包括任何评估芯片所共有的功能。 研究的第二阶段是采用一个案例研究中提出的解决方案，并使用第一个评估芯片设计期间开发的工具构建原型。 该项目的重点是将在一个问题领域有效的问题解决方法推广到更广泛的问题类别，并评估其在这些新领域的有效性。 该方法将这种强大的解决方案与现代 VLSI 技术相结合，产生独特的问题解决能力。