面向功能材料属性预测的机器学习方法初探
|
摘要
近年来,在功能材料领域通过手工方式从数百上千的候选结构中挖掘合理的功能材料结构.该方法过度依赖于经验知识,无法满足材料科学高效精准的结构预测.为了克服该问题,初步探究面向功能材料属性预测的机器学习方法,通过深入结合统计机器学习方法和预测功能材料结构,进而挖掘合理的影响材料结构属性的因素.具体而言,通过不同带隙建立材料数据,利用传统机器学习和深度学习算法搜索出可能的材料结构属性,进而利用DFT计算确定最优材料结构.从钙钛矿带隙数据和具有判别和表达能力的属性之间确定一组封闭的结构-性质关系映射,通过利用机器学习技术所学到的关系映射从精度和效率两个方面优化提升功能材料设计模式.
In recent years,reasonable functional material structures are usually excavated from hundreds of thousands of candidate structures in the field of functional materials by hand. Since this method relies too much on empirical knowledge,it fails to meet the material structure efficient and accurate structural prediction. In order to overcome this problem,this paper preliminarily explores the machine learning method for property prediction of functional materials. By deeply combining statistical machine learning methods and predicting functional material structure,this paper explores the factors that affect the material structure properties. Specifically,this paper establishes material data through different band gaps,and then uses traditional machine learning and deep learning algorithms to search for structure properties of possible materials. Then we use DFT calculation to determine the optimal material structure. The proposed methods determine a set of closed structure-property relationship mappings between perovskite bandgap data and attributes with discriminative and expressive capabilities. The relationship maps learned by machine learning techniques are optimized from both accuracy and efficiency,which promisingly improves the design patterns for functional materials.
In recent years,reasonable functional material structures are usually excavated from hundreds of thousands of candidate structures in the field of functional materials by hand. Since this method relies too much on empirical knowledge,it fails to meet the material structure efficient and accurate structural prediction. In order to overcome this problem,this paper preliminarily explores the machine learning method for property prediction of functional materials. By deeply combining statistical machine learning methods and predicting functional material structure,this paper explores the factors that affect the material structure properties. Specifically,this paper establishes material data through different band gaps,and then uses traditional machine learning and deep learning algorithms to search for structure properties of possible materials. Then we use DFT calculation to determine the optimal material structure. The proposed methods determine a set of closed structure-property relationship mappings between perovskite bandgap data and attributes with discriminative and expressive capabilities. The relationship maps learned by machine learning techniques are optimized from both accuracy and efficiency,which promisingly improves the design patterns for functional materials.
引文
[1]曾令可,罗民华,童晓濂,等.BP神经网络对陶瓷材料烧结性能预测的研究[J].材料导报,2002,(2):63-65.
[2]陈明和,谢兰生,朱知寿,等.计算机模拟与预测方法在材料科学研究中的应用[J].机械工程材料,2005,(6):1-3.
[3]李龙,徐松岭,杨惟高.BP神经网络在窑外分解窑优化控制中的应用[J].河北工业科技,2001,(2):35-37.
[4]余德泉.国内外工业机器人发展现状与趋势[J].大众用电,2017,(9):20-21.
[5]肖卓豪,卢安贤,刘树江,等.人工神经网络在玻璃配方设计中的应用研究[J].材料导报,2005,(6):17-19.
[6]Lowe D.G.Distinctive image features from scale invariant keypoints[J].International Journal of Computer Vision,2004,(60):91-110.
[7]Ojala T,Pietikainen M,Maenpaa T.Multiresolution grayscale and rotation invariant texture classification with local binary patterns[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2002,(7):971-987.
[8]Dalal N,Triggs B.Histograms of oriented gradients for human detection[A].Proceedings of International Conference on Computer Vision and Pattern Recognition[C].2005.
[9]Bengio Y,Lamblin P,Popovici D,et al.Greedy Layer-Wise Training of Deep Networks[A].Proceedings of Conference and Workshop on Neural Information Processing Systems[C].2006.
[10]Hinton G E,Osindero S,Teh Y W.A Fast Learning Algorithm for Deep Belief Nets[J].Neural Computation,2006,(7):1527-1554.
[11]Huang G,Lee H,Learned-miller E.Learning hierarchical representations for face verification with convolutional deep belief networks[A].Proceedings of International Conference on Computer Vision and Pattern Recognition[C].2012.
[12]Lu J,Liong V E,Zhou X,et al.Learning compact binary face descriptor for face recognition[J].IEEE Transactions on Pattern A-nalysis and Machine Intelligence,2015,(10):2041-2056.
[13]Le Q V,Karpenko A,Ngiam J.ICA with reconstruction cost for efficient overcomplete feature learning[A].Proceedings of Conference and Workshop on Neural Information Processing Systems[C].2011.
[14]Rifai S,Muller X,Glorot X.Contractive auto-encoders:explicit invariance during feature extraction[A].Proceedings of International Conference on Machine Learning[C].2011.
[15]Krizhevsky A,Sutskever I,Hinton G E.Imagenet classification with deep convolutional neural networks[A].Proceedings of Conference and Workshop on Neural Information Processing Systems[C].2012.
[16]Simonyan K,Zisserman A.Very deep convolutional networks for large-scale image recognition[A].Proceedings of International Conference on Learning Representation[C].2015.
[17]Ioffe S,Szegedy C.Batch normalization:accelerating deep network training by reducing internal covariate shift[A].Proceedings of International Conference on Machine Learning[C].2015.
[18]He K,Zhang X,Ren S,et al.Deep residual learning for image recognition[A].Proceedings of International Conference on Computer Vision and Pattern Recognition[C].2016.
[19]王正国,饶含兵,李泽荣.机器学习方法用于选择性环氧化酶-2抑制剂活性预测模型的建立[J].化学研究与应用,2006,(11):1317-1321.
[20]Sun Y,Bai H,Li Z,et al.Machine learning approach for prediction and understanding of glass-Forming ability[J].Journal Physics and Chemistry Letter,2017,(14):3434-3439.
[21]Hansen K,Biegler F,Ramakrishnan R,et al.Machine learning predictions of molecular properties:accurate many-body potentials and nonlocality in chemical space[J].Journal Physics and Chemistry Letter,2016,(12):2326-2331.
[22]Lu S,Zhou Q,Ouyang Y,et al.Accelerated discovery of stable lead-free hybrid organic-inorganic perovskites via machine learning[J].Nature Communications,2018,(9):3405.
[2]陈明和,谢兰生,朱知寿,等.计算机模拟与预测方法在材料科学研究中的应用[J].机械工程材料,2005,(6):1-3.
[3]李龙,徐松岭,杨惟高.BP神经网络在窑外分解窑优化控制中的应用[J].河北工业科技,2001,(2):35-37.
[4]余德泉.国内外工业机器人发展现状与趋势[J].大众用电,2017,(9):20-21.
[5]肖卓豪,卢安贤,刘树江,等.人工神经网络在玻璃配方设计中的应用研究[J].材料导报,2005,(6):17-19.
[6]Lowe D.G.Distinctive image features from scale invariant keypoints[J].International Journal of Computer Vision,2004,(60):91-110.
[7]Ojala T,Pietikainen M,Maenpaa T.Multiresolution grayscale and rotation invariant texture classification with local binary patterns[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2002,(7):971-987.
[8]Dalal N,Triggs B.Histograms of oriented gradients for human detection[A].Proceedings of International Conference on Computer Vision and Pattern Recognition[C].2005.
[9]Bengio Y,Lamblin P,Popovici D,et al.Greedy Layer-Wise Training of Deep Networks[A].Proceedings of Conference and Workshop on Neural Information Processing Systems[C].2006.
[10]Hinton G E,Osindero S,Teh Y W.A Fast Learning Algorithm for Deep Belief Nets[J].Neural Computation,2006,(7):1527-1554.
[11]Huang G,Lee H,Learned-miller E.Learning hierarchical representations for face verification with convolutional deep belief networks[A].Proceedings of International Conference on Computer Vision and Pattern Recognition[C].2012.
[12]Lu J,Liong V E,Zhou X,et al.Learning compact binary face descriptor for face recognition[J].IEEE Transactions on Pattern A-nalysis and Machine Intelligence,2015,(10):2041-2056.
[13]Le Q V,Karpenko A,Ngiam J.ICA with reconstruction cost for efficient overcomplete feature learning[A].Proceedings of Conference and Workshop on Neural Information Processing Systems[C].2011.
[14]Rifai S,Muller X,Glorot X.Contractive auto-encoders:explicit invariance during feature extraction[A].Proceedings of International Conference on Machine Learning[C].2011.
[15]Krizhevsky A,Sutskever I,Hinton G E.Imagenet classification with deep convolutional neural networks[A].Proceedings of Conference and Workshop on Neural Information Processing Systems[C].2012.
[16]Simonyan K,Zisserman A.Very deep convolutional networks for large-scale image recognition[A].Proceedings of International Conference on Learning Representation[C].2015.
[17]Ioffe S,Szegedy C.Batch normalization:accelerating deep network training by reducing internal covariate shift[A].Proceedings of International Conference on Machine Learning[C].2015.
[18]He K,Zhang X,Ren S,et al.Deep residual learning for image recognition[A].Proceedings of International Conference on Computer Vision and Pattern Recognition[C].2016.
[19]王正国,饶含兵,李泽荣.机器学习方法用于选择性环氧化酶-2抑制剂活性预测模型的建立[J].化学研究与应用,2006,(11):1317-1321.
[20]Sun Y,Bai H,Li Z,et al.Machine learning approach for prediction and understanding of glass-Forming ability[J].Journal Physics and Chemistry Letter,2017,(14):3434-3439.
[21]Hansen K,Biegler F,Ramakrishnan R,et al.Machine learning predictions of molecular properties:accurate many-body potentials and nonlocality in chemical space[J].Journal Physics and Chemistry Letter,2016,(12):2326-2331.
[22]Lu S,Zhou Q,Ouyang Y,et al.Accelerated discovery of stable lead-free hybrid organic-inorganic perovskites via machine learning[J].Nature Communications,2018,(9):3405.