透過您的圖書館登入
IP:3.141.202.187
  • 學位論文

時間序列預測在拉伸分子動力模擬之應用

Time-series Prediction in Steered Molecular Dynamics Simulation

指導教授 : 林太家

摘要


時間尺度上的限制在分子動力模擬的領域上一直以來都是一個 具有挑戰的問題。龐大的分子系統,例如:蛋白質,在超級電腦的 運算幫助下,仍然無法突破奈秒尺度的時間長度。這樣不完整的模 擬系統,會導致不完全的分子反應過程和結果。一個機器學習的技 法:時間序列預測,會在這篇論文中被使用,作為解決時間尺度問 題的方法。分子動力模擬中的不同變數,例如總能量,總力合,或 是部分的應力,甚至是氫鍵的強度,都能作為一筆資料來看待,而 這些資料,由於是分子模擬而來,會存在強烈的時間相依性,在這 樣強烈的時間相依性下,我們能將其視為時間序列來處理。利用數 據本身的特性,創造出兩筆不同時間軸的資料,利用彼此做訓練, 我們能透過機器學習的方法來預測這些數據的未來走向。

並列摘要


Time scale limitation has been and remains a challenging problem in Molecular Dynamic(MD) simulations. Large systems such as proteins are still narrowed in scale of nano-second with the help of super computers. The limit of time may contribute to insu cient result and improper inference while the simulation is already time-consuming. A machine learning method: time series prediction is implemented in this paper to solve this problem. Since features (Total force, numbers of Hydrogen bond, etc.) of protein have strong time dependence during simulation, we can use this relationship to obtain a different data by shifting original data with time. Taking these two data, we can train the machine to catch up structure of systems and achieve predicting these features in the next time step.

參考文獻


Fosang, Amanda J., et al. "Neutrophil collagenase (MMP-8) cleaves at the aggrecanase site E373A374 in the interglobular domain of cartilage aggrecan." Biochemical Journal 304.2 (1994): 347-351.
Fosang, Amanda J., et al. "Degradation of cartilage aggrecan by collagenase3 (MMP13)." FEBS letters 380.1-2 (1996): 17-20.
Gers, Felix A., Jrgen Schmidhuber, and Fred Cummins. "Learning to forget:Continual prediction with LSTM." (1999): 850-855.
Nakisa, Bahareh, et al. "Long short term memory hyperparameter optimization for a neural network based emotion recognition framework." IEEE Access 6 (2018): 49325-49338.
Gers, Felix A., Douglas Eck, and Jrgen Schmidhuber. "Applying LSTM to time series predictable through time-window approaches." Neural Nets WIRN Vietri-01. Springer, London, 2002. 193-200.

延伸閱讀