自我學習神經網路於多目標建築機能配置最適化之研究

The Optimal Study of Using Self-Learning Neural Network in Multi-Objective Architectural Function Layout

本研究以管理科學角度，將建築機能配置問題公式化為多目標組合最適化問題，並以「醫院診療大樓多目標機能配置」問題作個案應用。由於問題具多目標與組合爆炸的特性，為此本文先以二元型自我學習神經網路求解問題最佳解，再與隨機解作解答品質及求解效率比較。經測試得到本法解答的成果，遠優於隨機搜索的方法，以作為未來建築規劃相關人員，應用輔助參考之用。

This research is intended to formulate the problem of architectural function layout on the optimal problem of multi-objective combination and takes 'Multi-Objective Architectural Function Layout of Comprehensive Hospital Building' as a case study. Because of its multi-objective and combining explosive characteristics, the problem is solved by using 2-type self-learning neural network in this study. The random researching method is applied in resolving the problem and the results are also made comparison with the solution-quality and solving-efficiency. It is found that the testing results of self-learning neural network are superior to those of random researching method. The proposed result is useful for its application on architectural planning in the further.