遗传算法研究及化学结构信息处理
文献类型：学位论文
| 作者 | 何险峰 |
| 学位类别 | 博士 |
| 答辩日期 | 1999-12-24 |
| 授予单位 | 中国科学院研究生院 |
| 导师 | 周家驹 |
| 关键词 | 遗传标法 优化 化学结构信息处理 拓扑指数 联接矩陈 |
| 中文摘要 | 本论文针对分子设计中涉及的优化和化学结构信息处理问题进行了研究。遗传算法是近年发展起来的一类非数值优化算法。但经典遗传算法速度较慢、计算量较大、需要区分数据类型，针对这些缺点，本论文提出了改进的遗传算法IGA，IGA主要改进了遗传算法的算子和执行步骤，简化了染色体的编码过程，同时强调了杂交和变异的作用，使杂交和变异在演化群体的所有个体上执行。对经典测试函数的测试结果表明，IGA在性能上较经典遗传算法有较大的提高。结合均匀设计、有方向的优化及遗传演化的思想，本文提出了确定性遗传算法DGA，其基本思路是以空间上均匀分布的初始点展开确定性的演化以避免子空间的遗漏，以有方向性的杂交和变异加速和调节搜索，同时杂交和变异作用在不同层面上使搜索在整体上均匀化、通过遗传系统里的并行竞争提高整个群体的生存适应力，从而达到整体优休的目的。DGA不但对经典的优化测试函数表现出较强的优化能力，而且对一些非线性规划问题也表现出色。最后，基于对顺序树、联接矩阵和链烷烃结构式的观察，本文提出了一种能够反推结构式的双向拓扑指数——SOTI指数。SOTI指数的唯一性、不简并性和双向性使它对于化学结构数据库的存储、查询和管理大有益处。 |
| 英文摘要 | This thesis concerns with the problems related to the optimization and chemical structure information processing in computer-aided molecular design. The genetic algorithm (GA) is one kind of non-numeric optimization method. However, classical GAs have some disadvantages, such as low search speed, great amount of computation and requirement of differentiating data types. To these points, an improved genetic algorithm (IGA) proposed. The IGA focus on the improvement of genetic operators and executive procedure, which simplified the chromosome coding process and emphasized on the effects of crossover and mutation and made them perform on all individuals in population. It shows that the IGA gained more performance efficiency than the classical genetic algorithms by the validation on a set of classical test functions. Deterministic Genetic Algorithms (DGA) is another optimization algorithm, which, based on the acceptance of the ideas of uniform-design, directed optimization and genetic evolution, proposed in this thesis. The key points of DGA is to start deterministic evolution with initial population evenly distributed in searching space to avoid the leakiness of subspace, to accelerate the searching process by directed crossover and mutation, to improve holistic survival and adaptive ability by means of competition and parallel evolution in genetic system. However, the crossover and mutation act on different lays in space made the searching uniform. Not only to a set of classical test functions but also to some nonlinear programming problems does DGA show good optimization performance. Finally, inspired from the observation of the ordering-tree, connectivity matrix and structure of alkane, SOTI index, a two-direction topological with the functionality of deducing structure is proposed. However, uniqueness, low degeneracy and two-direction of SOTI index make it helpful to the storage, query and management of the structure database. |
| 语种 | 中文 |
| 公开日期 | 2013-09-27 |
| 页码 | 109 |
| 源URL | [http://ir.ipe.ac.cn/handle/122111/1993]  |
| 专题 | 过程工程研究所_研究所（批量导入） |
| 推荐引用方式 GB/T 7714 | 何险峰. 遗传算法研究及化学结构信息处理[D]. 中国科学院研究生院. 1999. |
入库方式： OAI收割
来源：过程工程研究所
浏览0
下载0
收藏0
其他版本
除非特别说明，本系统中所有内容都受版权保护，并保留所有权利。