Analyzing population diversity of modular program evolution through interactive visualization

这个研究项目解决了如何将遗传编程的机器学习技术的应用从小规模的例子转移到大规模的现实世界问题，比如开放式设计。工作假设是，一个能够组合模块化解决方案的学习系统可以更好地适应复杂问题。为了验证这一假设，我们开发了一个原型遗传编程系统，以模块形式产生解决方案。该系统在学习金融交易规则方面取得了一定的成功。本研究的目的是通过使用交互式可视化工具分析解决方案种群的多样性来了解系统的学习过程。有了这些知识，我们将能够有效地将该系统应用于大规模问题。

This research program addresses the question of how to move applications of the machine learning technique of genetic programming from small-scale examples to large-scale real world problems, such as open-ended design. The working hypothesis is that a learning system that is capable of composing modular solutions can better adapt to a complex problem. To verify this hypotheses, we have developed a prototype genetic programming system that produces solutions in modular form. The system has demonstrated some success in the task of learning financial trading rules. The objective of this research is to understand the learning process of the system by analyzing diversity of the solutions population using an interactive visualization tool. With that knowledge, we will be able to apply the system to large scale problems effectively.