Improving the Performance of AI Software: Payoffs and Pitfalls in Using Automatic Memoization
Improving the Performance of AI Software: Payoffs and Pitfalls in Using Automatic Memoization
复制标题
提高人工智能软件的性能：使用自动记忆的好处和陷阱
DOI：
--
复制
发表时间：
1993
期刊：
影响因子：
--
通讯作者：
J. Mayfield
中科院分区：
文献类型：
--
作者：
M. Hall;J. Mayfield
系统运行，并在离线会话期间节省昂贵的计算。即使在仅执行传统优化的应用程序中，它也是计时和分析代码的有用帮助。使用的简单性和透明性使该工具能够快速被程序员采用，并有助于最大限度地减少调试和验证时间。需要开发用于确定缓存值何时过期的工具。当将缓存内容存储到磁盘以供稍后会话使用时尤其如此。此外，关于记忆的不精确匹配，还有很多领域需要探索。 “模糊记忆”的有组织的框架可以为规划和/或学习中的应用提供实用性。最后，应开发限制备忘录表大小的方法（通过总条目或上次访问时间）。 6. 致谢 对代码和本文早期草稿提供宝贵反馈的人是 V.
a system run, and to save expensive calculations during an off-line session. It is also a useful aid for timing and profiling code, even in applications where only conventional optimization is being performed. The simplicity and transparency of use allows the facility to be quickly adopted by programmers, and helps minimize the debugging and verification time. Tools for determining when cached values are out of date need to be developed. This is particularly true when storing the contents of the cache to disk for use in a later session. Also, there is a lot of area to explore regarding inexact matches for memoization. An organized framework for " fuzzy memoization " might provide utility for applications in planning and/or learning. Finally, methods for limiting the size of the memo tables (either by total entries or time of last access) should be developed. 6. Acknowledgments People providing valuable feedback on the code and early drafts of the paper were V.