The good enough calculi of evolving control systems: evolution is not engineering.
The good enough calculi of evolving control systems: evolution is not engineering.
复制标题
进化控制系统的计算已经足够好了：进化不是工程。
DOI：
10.1152/ajpregu.1982.242.3.r173
复制
发表时间：
1982
期刊：
影响因子：
--
通讯作者：
L. Partridge
中科院分区：
文献类型：
--
作者：
L. Partridge
在偶然发明的元素的进化聚集体中，当统计学上足够好以识别前因系统的限制时，生存值可能有利于操作员结合反馈，前馈，状态变量和“稳态”控制的方面，而不是相同的。一般来说，简单的组织加薪应占主导地位。在发明之后，具有容易修改的规则的内部控制器可以促进复合发明的进化，但是控制规则变化的标准将仅是间接的（可能不完全的）生存指涉。与独立发明与间接标准和统计接受相结合，进化的控制逻辑可能是：冗余和不完整的;足够好的有害方面;建立松散的链接或自治的子逻辑;从不完整的信息足够好的解决方案的来源。部分明确的规则更多地是由拒绝而不是批准来定义的。基于形式逻辑、工程惯例和熟悉的坐标系对结果的研究可能会产生不合理的理解错觉。
In evolved aggregates of accidentally invented elements, retained when statistically good enough to identify limitations of antecedent systems, survival value might favor operators incorporating aspects of, while not identical with, feedback, feedforward, state varible, and "homeostatic" control. Generally, simple organizational increments should predominate. After invention, an internal controller with readily modifiable rules could facilitate evolution of compound inventions, but criteria controlling rule changes would be only indirectly (probably imperfectly) survival referent. Consequent to combination of independent invention with indirect criteria and statistical acceptance, evolved control logic could be: both redundant and incomplete; good enough with malefic aspects; built of loosely linked or autonomous sublogics; and a source of good enough solutions from incomplete information. The partially explicit rules are defined more by rejections than by ratifications. Study of the result based on formal logic, engineering conventions, and familiar coordinate systems could conceive illegitimate illusions of understanding.