Weihua Lei, Cleber Zanchettin, Flávio A. O. Santos, Luís A. Nunes Amaral
The extraordinary success of convolutional neural networks (CNNs) in various computer vision tasks has revitalized the field of artificial intelligence. The out-sized expectations created by this extraordinary success have, however, been tempered by a recognition of CNNs’ fragility. Importantly, the magnitude of the problem is unclear due to a lack of rigorous benchmark datasets. Here, we propose a solution to the benchmarking problem that reveals the extent of the vulnerabilities of CNNs and of the methods used to provide interpretability to their predictions. We employ cellular automata (CA) to generate images with rigorously controllable characteristics. CA allow for the definition of both extraordinarily simple and highly complex discrete functions and allow for the generation of boundless datasets of images without repeats. In this work, we systematically investigate the fragility and interpretability of the three popular CNN architectures using CA-generated datasets. We find a sharp transition from a learnable phase to an unlearnable phase as the latent space entropy of the discrete CA functions increases. Furthermore, we demonstrate that shortcut learning is an inherent trait of CNNs. Given a dataset with an easy-to-learn and strongly predictive pattern, CNN will consistently learn the shortcut even if the pattern occurs only on a small fraction of the image. Finally, we show that widely used attribution methods aiming to add interpretability to CNN outputs are strongly CNN-architecture specific and vary widely in their ability to identify input regions of high importance to the model. Our results provide significant insight into the limitations of both CNNs and the approaches developed to add interpretability to their predictions and raise concerns about the types of tasks that should be entrusted to them.
卷积神经网络（CNN）在各种计算机视觉任务中取得了非凡的成功，振兴了人工智能领域。然而，由于认识到卷积神经网络的脆弱性，人们对这一非凡成功产生了过高的期望。重要的是，由于缺乏严格的基准数据集，问题的严重性尚不明确。在此，我们提出了一个基准测试问题的解决方案，以揭示 CNN 的脆弱性程度，以及为其预测提供可解释性的方法。我们采用细胞自动机（CA）生成具有严格可控特征的图像。细胞自动机允许定义异常简单和高度复杂的离散函数，并允许生成无穷无尽的无重复图像数据集。在这项工作中，我们利用 CA 生成的数据集系统地研究了三种流行的 CNN 架构的脆弱性和可解释性。我们发现，随着离散 CA 函数的潜在空间熵的增加，可学习阶段会急剧过渡到不可学习阶段。此外，我们还证明了捷径学习是 CNN 的固有特性。如果数据集具有易于学习且预测性很强的模式，即使该模式只出现在一小部分图像上，CNN 也能持续学习该捷径。最后，我们表明，广泛使用的旨在为 CNN 输出增加可解释性的归因方法具有很强的 CNN 体系结构特性，在识别对模型非常重要的输入区域的能力方面存在很大差异。我们的研究结果让我们深入了解了 CNN 和为增加其预测的可解释性而开发的方法的局限性，并引起了人们对应该委托给它们的任务类型的关注。
{"title":"Computational experiments with cellular-automata generated images reveal intrinsic limitations of convolutional neural networks on pattern recognition tasks","authors":"Weihua Lei, Cleber Zanchettin, Flávio A. O. Santos, Luís A. Nunes Amaral","doi":"10.1063/5.0213905","DOIUrl":"https://doi.org/10.1063/5.0213905","url":null,"abstract":"The extraordinary success of convolutional neural networks (CNNs) in various computer vision tasks has revitalized the field of artificial intelligence. The out-sized expectations created by this extraordinary success have, however, been tempered by a recognition of CNNs’ fragility. Importantly, the magnitude of the problem is unclear due to a lack of rigorous benchmark datasets. Here, we propose a solution to the benchmarking problem that reveals the extent of the vulnerabilities of CNNs and of the methods used to provide interpretability to their predictions. We employ cellular automata (CA) to generate images with rigorously controllable characteristics. CA allow for the definition of both extraordinarily simple and highly complex discrete functions and allow for the generation of boundless datasets of images without repeats. In this work, we systematically investigate the fragility and interpretability of the three popular CNN architectures using CA-generated datasets. We find a sharp transition from a learnable phase to an unlearnable phase as the latent space entropy of the discrete CA functions increases. Furthermore, we demonstrate that shortcut learning is an inherent trait of CNNs. Given a dataset with an easy-to-learn and strongly predictive pattern, CNN will consistently learn the shortcut even if the pattern occurs only on a small fraction of the image. Finally, we show that widely used attribution methods aiming to add interpretability to CNN outputs are strongly CNN-architecture specific and vary widely in their ability to identify input regions of high importance to the model. Our results provide significant insight into the limitations of both CNNs and the approaches developed to add interpretability to their predictions and raise concerns about the types of tasks that should be entrusted to them.","PeriodicalId":502250,"journal":{"name":"APL Machine Learning","volume":"115 14","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-07-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141647072","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
A. McCray, Alec Bender, Amanda Petford-Long, C. Phatak
Understanding the collective behavior of complex spin textures, such as lattices of magnetic skyrmions, is of fundamental importance for exploring and controlling the emergent ordering of these spin textures and inducing phase transitions. It is also critical to understand the skyrmion–skyrmion interactions for applications such as magnetic skyrmion-enabled reservoir or neuromorphic computing. Magnetic skyrmion lattices can be studied using in situ Lorentz transmission electron microscopy (LTEM), but quantitative and statistically robust analysis of the skyrmion lattices from LTEM images can be difficult. In this work, we show that a convolutional neural network, trained on simulated data, can be applied to perform segmentation of spin textures and to extract quantitative data, such as spin texture size and location, from experimental LTEM images, which cannot be obtained manually. This includes quantitative information about skyrmion size, position, and shape, which can, in turn, be used to calculate skyrmion–skyrmion interactions and lattice ordering. We apply this approach to segmenting images of Néel skyrmion lattices so that we can accurately identify skyrmion size and deformation in both dense and sparse lattices. The model is trained using a large set of micromagnetic simulations as well as simulated LTEM images. This entirely open-source training pipeline can be applied to a wide variety of magnetic features and materials, enabling large-scale statistical studies of spin textures using LTEM.
{"title":"Simulation-trained machine learning models for Lorentz transmission electron microscopy","authors":"A. McCray, Alec Bender, Amanda Petford-Long, C. Phatak","doi":"10.1063/5.0197138","DOIUrl":"https://doi.org/10.1063/5.0197138","url":null,"abstract":"Understanding the collective behavior of complex spin textures, such as lattices of magnetic skyrmions, is of fundamental importance for exploring and controlling the emergent ordering of these spin textures and inducing phase transitions. It is also critical to understand the skyrmion–skyrmion interactions for applications such as magnetic skyrmion-enabled reservoir or neuromorphic computing. Magnetic skyrmion lattices can be studied using in situ Lorentz transmission electron microscopy (LTEM), but quantitative and statistically robust analysis of the skyrmion lattices from LTEM images can be difficult. In this work, we show that a convolutional neural network, trained on simulated data, can be applied to perform segmentation of spin textures and to extract quantitative data, such as spin texture size and location, from experimental LTEM images, which cannot be obtained manually. This includes quantitative information about skyrmion size, position, and shape, which can, in turn, be used to calculate skyrmion–skyrmion interactions and lattice ordering. We apply this approach to segmenting images of Néel skyrmion lattices so that we can accurately identify skyrmion size and deformation in both dense and sparse lattices. The model is trained using a large set of micromagnetic simulations as well as simulated LTEM images. This entirely open-source training pipeline can be applied to a wide variety of magnetic features and materials, enabling large-scale statistical studies of spin textures using LTEM.","PeriodicalId":502250,"journal":{"name":"APL Machine Learning","volume":"130 33","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141281750","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Recently, the rapid progress of deep learning techniques has brought unprecedented transformations and innovations across various fields. While neural network-based approaches can effectively encode data and detect underlying patterns of features, the diverse formats and compositions of data in different fields pose challenges in effectively utilizing these data, especially for certain research fields in the early stages of integrating deep learning. Therefore, it is crucial to find more efficient ways to utilize existing datasets. Here, we demonstrate that the predictive accuracy of the network can be improved dramatically by simply adding supplementary multi-frequency inputs to the existing dataset in the target spectrum predicting process. This design methodology paves the way for interdisciplinary research and applications at the interface of deep learning and other fields, such as photonics, composite material design, and biological medicine.
{"title":"Enhanced spectrum prediction using deep learning models with multi-frequency supplementary inputs","authors":"Xiaohua Xing, Yuqi Ren, Die Zou, Qiankun Zhang, Bingxuan Mao, Jianquan Yao, Deyi Xiong, Liang Wu","doi":"10.1063/5.0203931","DOIUrl":"https://doi.org/10.1063/5.0203931","url":null,"abstract":"Recently, the rapid progress of deep learning techniques has brought unprecedented transformations and innovations across various fields. While neural network-based approaches can effectively encode data and detect underlying patterns of features, the diverse formats and compositions of data in different fields pose challenges in effectively utilizing these data, especially for certain research fields in the early stages of integrating deep learning. Therefore, it is crucial to find more efficient ways to utilize existing datasets. Here, we demonstrate that the predictive accuracy of the network can be improved dramatically by simply adding supplementary multi-frequency inputs to the existing dataset in the target spectrum predicting process. This design methodology paves the way for interdisciplinary research and applications at the interface of deep learning and other fields, such as photonics, composite material design, and biological medicine.","PeriodicalId":502250,"journal":{"name":"APL Machine Learning","volume":"32 2","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-05-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"140967580","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Ziyao Zhang, Haoxiang Yang, J. K. Eshraghian, Jiayin Li, Ken-Tye Yong, D. Vigolo, Helen M. McGuire, Omid Kavehei
Imaging flow cytometry (IFC) is an advanced cell-analytic technology offering rich spatial information and fluorescence intensity for multi-parametric characterization. Manual gating in cytometry data enables the classification of discrete populations from the sample based on extracted features. However, this expert-driven technique can be subjective and laborious, often presenting challenges in reproducibility and being inherently limited to bivariate analysis. Numerous AI-driven cell classifications have recently emerged to automate the process of including multivariate data with enhanced reproducibility and accuracy. Our previous work demonstrated the early development of neuromorphic imaging cytometry, evaluating its feasibility in resolving conventional frame-based imaging systems’ limitations in data redundancy, fluorescence sensitivity, and compromised throughput. Herein, we adopted a convolutional spiking neural network (SNN) combined with the YOLOv3 model (SNN-YOLO) to perform cell classification and detection on label-free samples under neuromorphic vision. Spiking techniques are inherently suitable post-processing techniques for neuromorphic vision sensing. The experiment was conducted with polystyrene-based microparticles, THP-1, and LL/2 cell lines. The network’s performance was compared with that of a traditional YOLOv3 model fed with event-generated frame data to serve as a baseline. In this work, our SNN-YOLO outperformed the YOLOv3 baseline by achieving the highest average class accuracy of 0.974, compared to 0.962 for YOLOv3. Both models reported comparable performances across other key metrics and should be further explored for future auto-gating strategies and cytometry applications.
{"title":"Cell detection with convolutional spiking neural network for neuromorphic cytometry","authors":"Ziyao Zhang, Haoxiang Yang, J. K. Eshraghian, Jiayin Li, Ken-Tye Yong, D. Vigolo, Helen M. McGuire, Omid Kavehei","doi":"10.1063/5.0199514","DOIUrl":"https://doi.org/10.1063/5.0199514","url":null,"abstract":"Imaging flow cytometry (IFC) is an advanced cell-analytic technology offering rich spatial information and fluorescence intensity for multi-parametric characterization. Manual gating in cytometry data enables the classification of discrete populations from the sample based on extracted features. However, this expert-driven technique can be subjective and laborious, often presenting challenges in reproducibility and being inherently limited to bivariate analysis. Numerous AI-driven cell classifications have recently emerged to automate the process of including multivariate data with enhanced reproducibility and accuracy. Our previous work demonstrated the early development of neuromorphic imaging cytometry, evaluating its feasibility in resolving conventional frame-based imaging systems’ limitations in data redundancy, fluorescence sensitivity, and compromised throughput. Herein, we adopted a convolutional spiking neural network (SNN) combined with the YOLOv3 model (SNN-YOLO) to perform cell classification and detection on label-free samples under neuromorphic vision. Spiking techniques are inherently suitable post-processing techniques for neuromorphic vision sensing. The experiment was conducted with polystyrene-based microparticles, THP-1, and LL/2 cell lines. The network’s performance was compared with that of a traditional YOLOv3 model fed with event-generated frame data to serve as a baseline. In this work, our SNN-YOLO outperformed the YOLOv3 baseline by achieving the highest average class accuracy of 0.974, compared to 0.962 for YOLOv3. Both models reported comparable performances across other key metrics and should be further explored for future auto-gating strategies and cytometry applications.","PeriodicalId":502250,"journal":{"name":"APL Machine Learning","volume":" 10","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-05-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"140998464","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
J. Hinz, Dayou Yu, Deep Shankar Pandey, Hitesh Sapkota, Qi Yu, D. Mihaylov, V. V. Karasiev, S. Hu
Ab initio molecular dynamics (AIMD) simulations have become an important tool used in the construction of equations of state (EOS) tables for warm dense matter. Due to computational costs, only a limited number of system state conditions can be simulated, and the remaining EOS surface must be interpolated for use in radiation-hydrodynamic simulations of experiments. In this work, we develop a thermodynamically consistent EOS model that utilizes a physics-informed machine learning approach to implicitly learn the underlying Helmholtz free-energy from AIMD generated energies and pressures. The model, referred to as PIML-EOS, was trained and tested on warm dense polystyrene producing a fit within a 1% relative error for both energy and pressure and is shown to satisfy both the Maxwell and Gibbs–Duhem relations. In addition, we provide a path toward obtaining thermodynamic quantities, such as the total entropy and chemical potential (containing both ionic and electronic contributions), which are not available from current AIMD simulations.
原子分子动力学（AIMD）模拟已成为构建热致密物质状态方程（EOS）表的重要工具。由于计算成本的原因，只能模拟有限数量的系统状态条件，其余的 EOS 表必须内插到辐射流体力学模拟实验中使用。在这项工作中，我们开发了一种热力学一致的 EOS 模型，该模型利用物理信息机器学习方法，从 AIMD 生成的能量和压力中隐含地学习底层赫尔姆霍兹自由能。该模型被称为 PIML-EOS，在暖致密聚苯乙烯上进行了训练和测试，能量和压力的拟合相对误差均在 1%以内，并证明它同时满足麦克斯韦和吉布斯-杜恒关系。此外，我们还提供了一条获得热力学量的途径，如总熵和化学势（包含离子和电子贡献），这些都是目前的 AIMD 模拟无法获得的。
{"title":"The development of thermodynamically consistent and physics-informed equation-of-state model through machine learning","authors":"J. Hinz, Dayou Yu, Deep Shankar Pandey, Hitesh Sapkota, Qi Yu, D. Mihaylov, V. V. Karasiev, S. Hu","doi":"10.1063/5.0192447","DOIUrl":"https://doi.org/10.1063/5.0192447","url":null,"abstract":"Ab initio molecular dynamics (AIMD) simulations have become an important tool used in the construction of equations of state (EOS) tables for warm dense matter. Due to computational costs, only a limited number of system state conditions can be simulated, and the remaining EOS surface must be interpolated for use in radiation-hydrodynamic simulations of experiments. In this work, we develop a thermodynamically consistent EOS model that utilizes a physics-informed machine learning approach to implicitly learn the underlying Helmholtz free-energy from AIMD generated energies and pressures. The model, referred to as PIML-EOS, was trained and tested on warm dense polystyrene producing a fit within a 1% relative error for both energy and pressure and is shown to satisfy both the Maxwell and Gibbs–Duhem relations. In addition, we provide a path toward obtaining thermodynamic quantities, such as the total entropy and chemical potential (containing both ionic and electronic contributions), which are not available from current AIMD simulations.","PeriodicalId":502250,"journal":{"name":"APL Machine Learning","volume":"77 s327","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-05-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141002337","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
The flexibility of metal–organic frameworks (MOFs) affects their gas adsorption and diffusion properties. However, reliable force fields for simulating flexible MOFs are lacking. As a result, most atomistic simulations so far have been carried out assuming rigid MOFs, which inevitably overestimates the gas adsorption energy. Here, we show that this issue can be addressed by applying a machine-learning potential, trained on quantum chemistry data, to atomistic simulations. We find that inclusion of flexibility is particularly important for simulating CO2 chemisorption in MOFs with coordinatively unsaturated metal sites. Specifically, we demonstrate that the diffusion of CO2 in a flexible Mg-MOF-74 structure is about one order of magnitude faster than in a rigid one, challenging the rigid-MOF assumption in previous simulations.
{"title":"Simulating CO2 diffusivity in rigid and flexible Mg-MOF-74 with machine-learning force fields","authors":"Bowen Zheng, Grace X. Gu, Carine Ribeiro dos Santos, Rodrigo Neumann Barros Ferreira, Mathias Steiner, Binquan Luan","doi":"10.1063/5.0190372","DOIUrl":"https://doi.org/10.1063/5.0190372","url":null,"abstract":"The flexibility of metal–organic frameworks (MOFs) affects their gas adsorption and diffusion properties. However, reliable force fields for simulating flexible MOFs are lacking. As a result, most atomistic simulations so far have been carried out assuming rigid MOFs, which inevitably overestimates the gas adsorption energy. Here, we show that this issue can be addressed by applying a machine-learning potential, trained on quantum chemistry data, to atomistic simulations. We find that inclusion of flexibility is particularly important for simulating CO2 chemisorption in MOFs with coordinatively unsaturated metal sites. Specifically, we demonstrate that the diffusion of CO2 in a flexible Mg-MOF-74 structure is about one order of magnitude faster than in a rigid one, challenging the rigid-MOF assumption in previous simulations.","PeriodicalId":502250,"journal":{"name":"APL Machine Learning","volume":"96 7","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-05-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141003530","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Hu Quee Tan, Xinhai Zhao, Akhil Ambardekar, Erik Birgersson, Hansong Xue
Machine learning algorithms can enhance the design and experimental processing of solar cells, resulting in increased conversion efficiency. In this study, we introduce a novel machine learning-based methodology for optimizing the Pareto front of four-terminal (4T) perovskite-copper indium selenide (CIS) tandem solar cells (TSCs). By training a neural network using the Bayesian regularization-backpropagation algorithm via Hammersley sampling, we achieve high prediction accuracy when testing with unseen data through random sampling. This surrogate model not only reduces computational costs but also potentially enhances device performance, increasing from 29.4% to 30.4% while simultaneously reducing material costs for fabrication by 50%. Comparing experimentally fabricated cells with the predicted optimal cells, the latter show a thinner front contact electrode, charge-carrier transport layer, and back contact electrode. Highly efficient perovskite cells identified from the Pareto front have a perovskite layer thickness ranging from 420 to 580 nm. Further analysis reveals the front contact electrode needs to be thin, while the back contact electrode can have a thickness ranging from 100 to 145 nm and still achieve high efficiency. The charge-carrier transport layers play a crucial role in minimizing interface recombination and ensuring unidirectional current flow. The optimal design space suggests thinner electron and hole transport layer thicknesses of 7 nm, down from 23 to 10 nm, respectively. It indicates a balanced charge-carrier extraction is crucial for an optimized perovskite cell. Overall, the presented methodology and optimized design parameters have the potential to enhance the performance of 4T perovskite/CIS TSC while reducing material fabrication costs.
{"title":"Exploring the optimal design space of transparent perovskite solar cells for four-terminal tandem applications through Pareto front optimization","authors":"Hu Quee Tan, Xinhai Zhao, Akhil Ambardekar, Erik Birgersson, Hansong Xue","doi":"10.1063/5.0187208","DOIUrl":"https://doi.org/10.1063/5.0187208","url":null,"abstract":"Machine learning algorithms can enhance the design and experimental processing of solar cells, resulting in increased conversion efficiency. In this study, we introduce a novel machine learning-based methodology for optimizing the Pareto front of four-terminal (4T) perovskite-copper indium selenide (CIS) tandem solar cells (TSCs). By training a neural network using the Bayesian regularization-backpropagation algorithm via Hammersley sampling, we achieve high prediction accuracy when testing with unseen data through random sampling. This surrogate model not only reduces computational costs but also potentially enhances device performance, increasing from 29.4% to 30.4% while simultaneously reducing material costs for fabrication by 50%. Comparing experimentally fabricated cells with the predicted optimal cells, the latter show a thinner front contact electrode, charge-carrier transport layer, and back contact electrode. Highly efficient perovskite cells identified from the Pareto front have a perovskite layer thickness ranging from 420 to 580 nm. Further analysis reveals the front contact electrode needs to be thin, while the back contact electrode can have a thickness ranging from 100 to 145 nm and still achieve high efficiency. The charge-carrier transport layers play a crucial role in minimizing interface recombination and ensuring unidirectional current flow. The optimal design space suggests thinner electron and hole transport layer thicknesses of 7 nm, down from 23 to 10 nm, respectively. It indicates a balanced charge-carrier extraction is crucial for an optimized perovskite cell. Overall, the presented methodology and optimized design parameters have the potential to enhance the performance of 4T perovskite/CIS TSC while reducing material fabrication costs.","PeriodicalId":502250,"journal":{"name":"APL Machine Learning","volume":"100 4","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-04-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"140659225","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Jiaxiang Chen, Haitao Du, Haolan Qu, Han Gao, Yitian Gu, Yitai Zhu, Wenbo Ye, Jun Zou, Hongzhi Wang, Xinbo Zou
Artificial optoelectronic synaptic transistors have attracted extensive research interest as an essential component for neuromorphic computing systems and brain emulation applications. However, performance challenges still remain for synaptic devices, including low energy consumption, high integration density, and flexible modulation. Employing trapping and detrapping relaxation, a novel optically stimulated synaptic transistor enabled by the AlGaN/GaN hetero-structure metal-oxide semiconductor high-electron-mobility transistor has been successfully demonstrated in this study. Synaptic functions, including excitatory postsynaptic current (EPSC), paired-pulse facilitation index, and transition from short-term memory to long-term memory, are well mimicked and explicitly investigated. In a single EPSC event, the AlGaN/GaN synaptic transistor shows the characteristics of low energy consumption and a high signal-to-noise ratio. The EPSC of the synaptic transistor can be synergistically modulated by both optical stimulation and gate/drain bias. Moreover, utilizing a convolution neural network, hand-written digit images were used to verify the data preprocessing capability for neuromorphic computing applications.
{"title":"AlGaN/GaN MOS-HEMT enabled optoelectronic artificial synaptic devices for neuromorphic computing","authors":"Jiaxiang Chen, Haitao Du, Haolan Qu, Han Gao, Yitian Gu, Yitai Zhu, Wenbo Ye, Jun Zou, Hongzhi Wang, Xinbo Zou","doi":"10.1063/5.0194083","DOIUrl":"https://doi.org/10.1063/5.0194083","url":null,"abstract":"Artificial optoelectronic synaptic transistors have attracted extensive research interest as an essential component for neuromorphic computing systems and brain emulation applications. However, performance challenges still remain for synaptic devices, including low energy consumption, high integration density, and flexible modulation. Employing trapping and detrapping relaxation, a novel optically stimulated synaptic transistor enabled by the AlGaN/GaN hetero-structure metal-oxide semiconductor high-electron-mobility transistor has been successfully demonstrated in this study. Synaptic functions, including excitatory postsynaptic current (EPSC), paired-pulse facilitation index, and transition from short-term memory to long-term memory, are well mimicked and explicitly investigated. In a single EPSC event, the AlGaN/GaN synaptic transistor shows the characteristics of low energy consumption and a high signal-to-noise ratio. The EPSC of the synaptic transistor can be synergistically modulated by both optical stimulation and gate/drain bias. Moreover, utilizing a convolution neural network, hand-written digit images were used to verify the data preprocessing capability for neuromorphic computing applications.","PeriodicalId":502250,"journal":{"name":"APL Machine Learning","volume":"51 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-04-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"140664450","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Knowledge of condensate shedding droplet dynamics provides important information for the characterization of two-phase heat and mass transfer phenomena. Detecting and segmenting the droplets during shedding requires considerable time and effort if performed manually. Here, we developed a self-supervised deep learning model for segmenting shedding droplets from a variety of dropwise and filmwise condensing surfaces. The model eliminates the need for image annotation by humans in the training step and, therefore, reduces labor significantly. The trained model achieved an average accuracy greater than 0.9 on a new unseen test dataset. After extracting the shedding droplet size and speed, we developed a data-driven model for shedding droplet dynamics based on condensation heat flux and surface properties such as wettability and tube diameter. Our results demonstrate that condensate droplet departure size is both heat flux and tube size dependent and follows different trends based on the condensation mode. The results of this work provide an annotation-free methodology for falling droplet segmentation as well as a statistical understanding of droplet dynamics during condensation.
{"title":"Self-supervised learning of shedding droplet dynamics during steam condensation","authors":"Siavash Khodakarami, Pouya Kabirzadeh, Nenad Miljkovic","doi":"10.1063/5.0188620","DOIUrl":"https://doi.org/10.1063/5.0188620","url":null,"abstract":"Knowledge of condensate shedding droplet dynamics provides important information for the characterization of two-phase heat and mass transfer phenomena. Detecting and segmenting the droplets during shedding requires considerable time and effort if performed manually. Here, we developed a self-supervised deep learning model for segmenting shedding droplets from a variety of dropwise and filmwise condensing surfaces. The model eliminates the need for image annotation by humans in the training step and, therefore, reduces labor significantly. The trained model achieved an average accuracy greater than 0.9 on a new unseen test dataset. After extracting the shedding droplet size and speed, we developed a data-driven model for shedding droplet dynamics based on condensation heat flux and surface properties such as wettability and tube diameter. Our results demonstrate that condensate droplet departure size is both heat flux and tube size dependent and follows different trends based on the condensation mode. The results of this work provide an annotation-free methodology for falling droplet segmentation as well as a statistical understanding of droplet dynamics during condensation.","PeriodicalId":502250,"journal":{"name":"APL Machine Learning","volume":"156 20 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-04-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"140717731","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Refractory non-dilute random alloys consist of two or more principal refractory metals with complex interactions that modify their basic structural properties such as lattice parameters and elastic constants. Atomistic simulations (ASs) are an effective method to compute such basic structural parameters. However, accurate predictions from ASs are computationally expensive due to the size and number of atomistic structures required. To reduce the computational burden, multivariate Gaussian process regression (MVGPR) is proposed as a surrogate model that only requires computing a small number of configurations for training. The elemental atom percentage in the hyper-spherical coordinates is demonstrated to be an effective feature for surrogate modeling. An additive approximation of the full MVGPR model is also proposed to further reduce computations. To improve surrogate accuracy, active learning is used to select a small number of alloys to simulate. Numerical studies based on AS data show the accuracy of the surrogate methodology and the additive approximation, as well as the effectiveness and robustness of the active learning for selecting new alloy designs to simulate.
难熔非稀释无规合金由两种或两种以上的主要难熔金属组成，其复杂的相互作用改变了它们的基本结构特性，如晶格参数和弹性常数。原子模拟（AS）是计算此类基本结构参数的有效方法。然而，由于需要原子结构的大小和数量，通过原子模拟进行精确预测的计算成本很高。为了减轻计算负担，我们提出了多变量高斯过程回归（MVGPR）作为一种替代模型，它只需要计算少量的构型进行训练。超球面坐标中的元素原子百分比被证明是代用模型的有效特征。此外，还提出了完整 MVGPR 模型的加法近似值，以进一步减少计算量。为了提高代用精度，采用了主动学习方法来选择少量合金进行模拟。基于 AS 数据的数值研究显示了代用方法和加法近似的准确性，以及主动学习在选择新合金设计进行模拟时的有效性和稳健性。
{"title":"Multivariate Gaussian process surrogates for predicting basic structural parameters of refractory non-dilute random alloys","authors":"Cesar Ruiz, Anshu Raj, Shuozhi Xu","doi":"10.1063/5.0186045","DOIUrl":"https://doi.org/10.1063/5.0186045","url":null,"abstract":"Refractory non-dilute random alloys consist of two or more principal refractory metals with complex interactions that modify their basic structural properties such as lattice parameters and elastic constants. Atomistic simulations (ASs) are an effective method to compute such basic structural parameters. However, accurate predictions from ASs are computationally expensive due to the size and number of atomistic structures required. To reduce the computational burden, multivariate Gaussian process regression (MVGPR) is proposed as a surrogate model that only requires computing a small number of configurations for training. The elemental atom percentage in the hyper-spherical coordinates is demonstrated to be an effective feature for surrogate modeling. An additive approximation of the full MVGPR model is also proposed to further reduce computations. To improve surrogate accuracy, active learning is used to select a small number of alloys to simulate. Numerical studies based on AS data show the accuracy of the surrogate methodology and the additive approximation, as well as the effectiveness and robustness of the active learning for selecting new alloy designs to simulate.","PeriodicalId":502250,"journal":{"name":"APL Machine Learning","volume":"104 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-04-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"140720480","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：