演化過程之視覺化

Abstract

基因演算法在許多領域中已獲得不錯的成效，但我們卻缺乏一個適當方法來觀察演化的進行。這是由 於基因演算法運算子之運作具隨機性，使我們不易掌握演化的機制。有鑑於此，本論文嘗試使用視覺化方 法來解決此一問題。 在本論文中，我們結合性擇模型和視覺化的技巧，將性擇模型之模擬實驗數據視覺化，並加以分析， 希望藉此表現出隱藏在演化中之整體趨勢以及隱藏之形態。 本論文以性擇模型為基礎，在文中我們提出了五種視覺化的方法，分別從族群及基因的角度來觀察演 化的進行。藉由這些視覺化方法，我們觀察到一些原本從記錄資料中不易發現的資訊。最後，我們在論文 中討論這些方法的特性及其優缺點。

Although Genetic Algorithms have been performing well in many fields, we still need a proper method to investigate the evolution processes in more detail. Due to the stochastic nature of the genetic operators, we are not able to mathematically analyze the mechanism of evolution. In this thesis we try to provide a solution by employing the concept of data visualization. In this study, we apply visualization technologies to visualize the data logged from the simulation of a sexual selection model in order to find the global trends and hidden patterns in the evolutionary process. We propose five visualization methods based on the sexual selection model from both the population and the gene points of view. Some important pieces of information were unveiled via these visualization methods. We also discuss the advantages and disadvantages of those five visualization methods in this thesis.