Yuan-Hong Tsai
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Yuan-Hong Tsai has a total of 16 co-authors. Here are the top co-authors by number of shared publications:
- Ming-Chiang Chung
Department of Physics, National Chung Hsing University 9 , Taichung
9 shared publications Guang-Yu Huang
Physics Department, National Chung-Hsing University, Taichung 40227, Taiwan, Republic of China
6 shared publicationsTsung-Pao Cheng
Physics Department, National Chung-Hsing University, Taichung 40227, Taiwan
4 shared publicationsKuan-Jung Su
Physics Department, National Chung-Hsing University, Taichung 40227, Taiwan
2 shared publicationsKuo-Feng Chiu
Physics Department, National Chung-Hsing University, Taichung 40227, Taiwan
2 shared publicationsTzu-Pei Yang
Physics Department, National Chung-Hsing University, Taichung 40227, Taiwan
2 shared publications
Latest Publications
- Open Access
- 2 July 2026
- Published by IOP Publishing in Machine Learning: Science and Technology
- Vol. 7 (4) , 045003
Machine learning framework for GW calculations across molecular dynamics trajectories
Machine learning offers a transformative approach to overcoming the computational bottlenecks of quasiparticle energy calculations in many-body perturbation theory. While the G 0 W 0 approximation provides a rigorous description of excited-state electronic properties, its high computational cost—driven by the evaluation of the inverse dielectric matrix—renders it prohibitive for molecular dynamics (MD) simulations that require sampling thousands of thermal configurations. We present a gradient-boosting framework (LightGBM) that predicts G 0 W 0 quasiparticle energies with high efficiency using only computationally inexpensive mean-field density functional theory eigenvalues and exchange-correlation potentials. Trained on a minimal subset (25%) of MD snapshots for silicon and boron nitride, the model achieves root-mean-square errors below 0.1 eV, accurately reproducing full k -resolved band structures and density of states. Crucially, we demonstrate the model’s transferability and robustness: it successfully generalizes to unseen simulation times, different thermal regimes, and distinct crystal polymorphs excluded from the training data. This data-driven approach effectively bypasses the explicit calculation of the self-energy operator, offering a scalable pathway for high-throughput excited-state simulations and the study of finite-temperature electronic properties in complex materials.
- 2 July 2026
- Published by IOP Publishing in Machine Learning: Science and Technology
- Vol. 7 (4) , 045003
Open AccessMachine learning framework for GW calculations across molecular dynamics trajectories
Machine learning offers a transformative approach to overcoming the computational bottlenecks of quasiparticle energy calculations in many-body perturbation theory. While the G 0 W 0 approximation provides a rigorous description of excited-state electronic properties, its high computational cost—driven by the evaluation of the inverse dielectric matrix—renders it prohibitive for molecular dynamics (MD) simulations that require sampling thousands of thermal configurations. We present a gradient-boosting framework (LightGBM) that predicts G 0 W 0 quasiparticle energies with high efficiency using only computationally inexpensive mean-field density functional theory eigenvalues and exchange-correlation potentials. Trained on a minimal subset (25%) of MD snapshots for silicon and boron nitride, the model achieves root-mean-square errors below 0.1 eV, accurately reproducing full k -resolved band structures and density of states. Crucially, we demonstrate the model’s transferability and robustness: it successfully generalizes to unseen simulation times, different thermal regimes, and distinct crystal polymorphs excluded from the training data. This data-driven approach effectively bypasses the explicit calculation of the self-energy operator, offering a scalable pathway for high-throughput excited-state simulations and the study of finite-temperature electronic properties in complex materials.
- Open Access
- 29 June 2023
- Published by American Physical Society (APS) in Physical Review B
- Vol. 107 (21) , 214451
Deep learning of phase transitions for quantum spin chains from correlation aspects
Using machine learning (ML) to recognize different phases of matter and to infer the entire phase diagram has proven to be an effective tool given a large dataset. In our previous proposals, we have successfully explored phase transitions for topological phases of matter at low dimensions either in a supervised or an unsupervised learning protocol with the assistance of quantum-information-related quantities. In this work, we adopt our previous ML procedures to study quantum phase transitions of magnetism systems such as the and spin chains by using spin-spin correlation functions as the input data. We find that our proposed approach not only maps out the phase diagrams with accurate phase boundaries, but also indicates some features that have not been observed in the field of machine learning before. In particular, we define so-called relevant correlation functions to some corresponding phases that can always distinguish between those and their neighbors. Based on the unsupervised learning protocol we proposed [ Phys. Rev. B 104, 165108 (2021)], the reduced latent representations of the inputs combined with the clustering algorithm show the connectedness or disconnectedness between neighboring clusters (phases) just corresponding to the continuous or disrupt quantum phase transition, respectively. This property reminds us of the behavior of order parameters. Moreover, in the silhouette analysis we show that the ferromagnetic states in the model with various anisotropy parameters correspond to almost the same silhouette value, while the critical or antiferromagnetic states behave quite differently. The analysis further indicates that the minima of silhouette values are close to the phase-transition points, showing strong positive correlation. These results again justify the usefulness of our proposed ML procedures, and they move us a step further toward understanding the relation between ML and quantum phase transitions from correlation function aspects.- 29 June 2023
- Published by American Physical Society (APS) in Physical Review B
- Vol. 107 (21) , 214451
Open AccessDeep learning of phase transitions for quantum spin chains from correlation aspects
Using machine learning (ML) to recognize different phases of matter and to infer the entire phase diagram has proven to be an effective tool given a large dataset. In our previous proposals, we have successfully explored phase transitions for topological phases of matter at low dimensions either in a supervised or an unsupervised learning protocol with the assistance of quantum-information-related quantities. In this work, we adopt our previous ML procedures to study quantum phase transitions of magnetism systems such as the and spin chains by using spin-spin correlation functions as the input data. We find that our proposed approach not only maps out the phase diagrams with accurate phase boundaries, but also indicates some features that have not been observed in the field of machine learning before. In particular, we define so-called relevant correlation functions to some corresponding phases that can always distinguish between those and their neighbors. Based on the unsupervised learning protocol we proposed [ Phys. Rev. B 104, 165108 (2021)], the reduced latent representations of the inputs combined with the clustering algorithm show the connectedness or disconnectedness between neighboring clusters (phases) just corresponding to the continuous or disrupt quantum phase transition, respectively. This property reminds us of the behavior of order parameters. Moreover, in the silhouette analysis we show that the ferromagnetic states in the model with various anisotropy parameters correspond to almost the same silhouette value, while the critical or antiferromagnetic states behave quite differently. The analysis further indicates that the minima of silhouette values are close to the phase-transition points, showing strong positive correlation. These results again justify the usefulness of our proposed ML procedures, and they move us a step further toward understanding the relation between ML and quantum phase transitions from correlation function aspects. - Open Access
Deep Learning of Phase Transitions for Quantum Spin Chains from Correlation Aspects
Using machine learning (ML) to recognize different phases of matter and to infer the entire phase diagram has proven to be an effective tool given a large dataset. In our previous proposals, we have successfully explored phase transitions for topological phases of matter at low dimensions either in a supervised or an unsupervised learning protocol with the assistance of quantum information related quantities. In this work, we adopt our previous ML procedures to study quantum phase transitions of magnetism systems such as the XY and XXZ spin chains by using spin-spin correlation functions as the input data. We find that our proposed approach not only maps out the phase diagrams with accurate phase boundaries, but also indicates some new features that have not observed before. In particular, we define so-called relevant correlation functions to some corresponding phases that can always distinguish between those and their neighbors. Based on the unsupervised learning protocol we proposed [Phys. Rev. B 104, 165108 (2021)], the reduced latent representations of the inputs combined with the clustering algorithm show the connectedness or disconnectedness between neighboring clusters (phases), just corresponding to the continuous or disrupt quantum phase transition, respectively.
Open AccessDeep Learning of Phase Transitions for Quantum Spin Chains from Correlation Aspects
Using machine learning (ML) to recognize different phases of matter and to infer the entire phase diagram has proven to be an effective tool given a large dataset. In our previous proposals, we have successfully explored phase transitions for topological phases of matter at low dimensions either in a supervised or an unsupervised learning protocol with the assistance of quantum information related quantities. In this work, we adopt our previous ML procedures to study quantum phase transitions of magnetism systems such as the XY and XXZ spin chains by using spin-spin correlation functions as the input data. We find that our proposed approach not only maps out the phase diagrams with accurate phase boundaries, but also indicates some new features that have not observed before. In particular, we define so-called relevant correlation functions to some corresponding phases that can always distinguish between those and their neighbors. Based on the unsupervised learning protocol we proposed [Phys. Rev. B 104, 165108 (2021)], the reduced latent representations of the inputs combined with the clustering algorithm show the connectedness or disconnectedness between neighboring clusters (phases), just corresponding to the continuous or disrupt quantum phase transition, respectively.
- 5 October 2021
- Published by American Physical Society (APS) in Physical Review B
- Vol. 104 (16) , 165108
Deep learning of topological phase transitions from entanglement aspects: An unsupervised way
Machine learning techniques have been shown to be effective to recognize different phases of matter and produce phase diagrams in the parameter space of interest, while they usually require prior labeled data to perform well. Here, we propose a machine learning procedure, mainly in an unsupervised manner, which can first identify topological/nontopological phases and then refine the locations of phase boundaries. By following this proposed procedure, we expand our previous work on the one-dimensional -wave superconductor [Tsai et al., Phys. Rev. B 102, 054512 (2020)] and further on the Su-Schrieffer-Heeger model, with an emphasis on using the quantum-entanglement-based quantities as the input features. We find that our method not only reproduces similar results to the previous work with sharp phase boundaries but importantly it also does not rely on prior knowledge of the phase space, e.g., the number of phases present. We conclude with a few remarks about its potential, limitations, and explanatory power.- 5 October 2021
- Published by American Physical Society (APS) in Physical Review B
- Vol. 104 (16) , 165108
Deep learning of topological phase transitions from entanglement aspects: An unsupervised way
Machine learning techniques have been shown to be effective to recognize different phases of matter and produce phase diagrams in the parameter space of interest, while they usually require prior labeled data to perform well. Here, we propose a machine learning procedure, mainly in an unsupervised manner, which can first identify topological/nontopological phases and then refine the locations of phase boundaries. By following this proposed procedure, we expand our previous work on the one-dimensional -wave superconductor [Tsai et al., Phys. Rev. B 102, 054512 (2020)] and further on the Su-Schrieffer-Heeger model, with an emphasis on using the quantum-entanglement-based quantities as the input features. We find that our method not only reproduces similar results to the previous work with sharp phase boundaries but importantly it also does not rely on prior knowledge of the phase space, e.g., the number of phases present. We conclude with a few remarks about its potential, limitations, and explanatory power.- Open Access
Deep learning of topological phase transitions from entanglement aspects for two-dimensional chiral p-wave superconductors
Applying deep learning to investigate topological phase transitions (TPTs) becomes a useful method due to not only its ability to recognize patterns but also its statistical excellency to examine the amount of information carried by different types of data inputs. Among possible data types, entanglement-related quantities, such as Majorana correlation matrices (MCMs), one-particle entanglement spectra (OPES), and entanglement eigenvectors (OPEEs), have been proved effective, however, are to date mostly restricted to one dimension. Here, we propose practical input data forms based on those quantities to study TPTs and to compare the efficiency of each form on classic two-dimensional chiral $p$-wave superconductors via the deep learning approach. First, we find that different input forms, either matrices or tensors both originated from real MCMs, can affect the precise locations of the predicted transition points. Next, due to the complex nature of OPEEs, we extract three spatially dependent quantities from OPEEs, one related to the "intensity", and the other two related to "phases" of particle and hole components. We show that similar to taking OPES directly as inputs, solely using "intensity" quantity can only distinguish topological phases from trivial ones, whereas using either whole MCMs or complete OPEE-extracted quantities can provide sufficient information for deep learning to distinguish between phases of matter with different $U(1)$ gauges or Chern numbers. Finally, we discuss certain characteristic features in the deep learning approach and, in particular, they reveal that our trained models indeed learn physically meaningful features, which confirms the potential use even at high dimensions.Open AccessDeep learning of topological phase transitions from entanglement aspects for two-dimensional chiral p-wave superconductors
Applying deep learning to investigate topological phase transitions (TPTs) becomes a useful method due to not only its ability to recognize patterns but also its statistical excellency to examine the amount of information carried by different types of data inputs. Among possible data types, entanglement-related quantities, such as Majorana correlation matrices (MCMs), one-particle entanglement spectra (OPES), and entanglement eigenvectors (OPEEs), have been proved effective, however, are to date mostly restricted to one dimension. Here, we propose practical input data forms based on those quantities to study TPTs and to compare the efficiency of each form on classic two-dimensional chiral $p$-wave superconductors via the deep learning approach. First, we find that different input forms, either matrices or tensors both originated from real MCMs, can affect the precise locations of the predicted transition points. Next, due to the complex nature of OPEEs, we extract three spatially dependent quantities from OPEEs, one related to the "intensity", and the other two related to "phases" of particle and hole components. We show that similar to taking OPES directly as inputs, solely using "intensity" quantity can only distinguish topological phases from trivial ones, whereas using either whole MCMs or complete OPEE-extracted quantities can provide sufficient information for deep learning to distinguish between phases of matter with different $U(1)$ gauges or Chern numbers. Finally, we discuss certain characteristic features in the deep learning approach and, in particular, they reveal that our trained models indeed learn physically meaningful features, which confirms the potential use even at high dimensions. - 22 July 2021
- Published by American Physical Society (APS) in Physical Review B
- Vol. 104 (2) , 024506
Deep learning of topological phase transitions from the point of view of entanglement for two-dimensional chiral -wave superconductorsApplying deep learning to investigate topological phase transitions (TPTs) becomes a useful method due to not only its ability to recognize patterns but also its statistical excellency to examine the amount of information carried by different types of data inputs. Among possible data types, entanglement-related quantities, such as Majorana correlation matrices (MCMs), one-particle entanglement spectra (OPES), and entanglement eigenvectors (OPEEs), have been proved effective, however, are to date mostly restricted to one dimension. Here, we propose practical input data forms based on those quantities to study TPTs and to compare the efficiency of each form on classic two-dimensional chiral -wave superconductors via the deep learning approach. First, we find that different input forms, either matrices or tensors both originated from real MCMs, can affect the precise locations of the predicted transition points. Next, due to the complex nature of OPEEs, we extract three spatially dependent quantities from OPEEs, one related to the “intensity,” and the other two related to “phases” of particle and hole components. We show that similar to taking OPES directly as inputs, solely using “intensity” quantity can only distinguish topological phases from trivial ones, whereas using either whole MCMs or complete OPEE-extracted quantities can provide sufficient information for deep learning to distinguish between phases of matter with different gauges or Chern numbers. Finally, we discuss certain characteristic features in the deep learning approach and, in particular, they reveal that our trained models indeed learn physically meaningful features, which confirms the potential use even at high dimensions.- 22 July 2021
- Published by American Physical Society (APS) in Physical Review B
- Vol. 104 (2) , 024506
Deep learning of topological phase transitions from the point of view of entanglement for two-dimensional chiral -wave superconductorsApplying deep learning to investigate topological phase transitions (TPTs) becomes a useful method due to not only its ability to recognize patterns but also its statistical excellency to examine the amount of information carried by different types of data inputs. Among possible data types, entanglement-related quantities, such as Majorana correlation matrices (MCMs), one-particle entanglement spectra (OPES), and entanglement eigenvectors (OPEEs), have been proved effective, however, are to date mostly restricted to one dimension. Here, we propose practical input data forms based on those quantities to study TPTs and to compare the efficiency of each form on classic two-dimensional chiral -wave superconductors via the deep learning approach. First, we find that different input forms, either matrices or tensors both originated from real MCMs, can affect the precise locations of the predicted transition points. Next, due to the complex nature of OPEEs, we extract three spatially dependent quantities from OPEEs, one related to the “intensity,” and the other two related to “phases” of particle and hole components. We show that similar to taking OPES directly as inputs, solely using “intensity” quantity can only distinguish topological phases from trivial ones, whereas using either whole MCMs or complete OPEE-extracted quantities can provide sufficient information for deep learning to distinguish between phases of matter with different gauges or Chern numbers. Finally, we discuss certain characteristic features in the deep learning approach and, in particular, they reveal that our trained models indeed learn physically meaningful features, which confirms the potential use even at high dimensions.- Open Access
Deep learning of topological phase transitions from entanglement aspects: An unsupervised way
Machine learning techniques have been shown to be effective to recognize different phases of matter and produce phase diagrams in the parameter space interested, while they usually require prior labeled data to perform well. Here, we propose a machine learning procedure, mainly in an unsupervised manner, which can first identify topological/non-topological phases and then refine the locations of phase boundaries. By following this proposed procedure, we expand our previous work on the one-dimensional $p$-wave superconductor [Phys. Rev. B 102, 054512 (2020)] and further on the Su-Schrieffer-Heeger model, with an emphasis on using the quantum entanglement-based quantities as the input features. We find that our method not only reproduces similar results to the previous work with sharp phase boundaries but importantly it also does not rely on prior knowledge of the phase space, e.g., the number of phases present. We conclude with a few remarks about its potential, limitations, and explainabilities.
Open AccessDeep learning of topological phase transitions from entanglement aspects: An unsupervised way
Machine learning techniques have been shown to be effective to recognize different phases of matter and produce phase diagrams in the parameter space interested, while they usually require prior labeled data to perform well. Here, we propose a machine learning procedure, mainly in an unsupervised manner, which can first identify topological/non-topological phases and then refine the locations of phase boundaries. By following this proposed procedure, we expand our previous work on the one-dimensional $p$-wave superconductor [Phys. Rev. B 102, 054512 (2020)] and further on the Su-Schrieffer-Heeger model, with an emphasis on using the quantum entanglement-based quantities as the input features. We find that our method not only reproduces similar results to the previous work with sharp phase boundaries but importantly it also does not rely on prior knowledge of the phase space, e.g., the number of phases present. We conclude with a few remarks about its potential, limitations, and explainabilities.
- Open Access
- 18 August 2020
- Published by American Physical Society (APS) in Physical Review B
- Vol. 102 (5) , 054512
Deep learning of topological phase transitions from entanglement aspects
The one-dimensional -wave superconductor proposed by Kitaev has long been a classic example for understanding topological phase transitions through various methods, such as examining the Berry phase, edge states of open chains, and, in particular, aspects from quantum entanglement of ground states. In order to understand the amount of information carried in the entanglement-related quantities, here we study topological phase transitions of the model with emphasis of using the deep learning approach. We feed different quantities, including Majorana correlation matrices (MCMs), entanglement spectra (ES) or entanglement eigenvectors (EE) originating from Block correlation matrices, into the deep neural networks for training, and investigate which one could be the most useful input format in this approach. We find that ES is information that is too compressed compared to MCM or EE. MCM and EE can provide us abundant information to recognize not only the topological phase transitions in the model but also phases of matter with different (1) gauges, which is not reachable by using ES only.- 18 August 2020
- Published by American Physical Society (APS) in Physical Review B
- Vol. 102 (5) , 054512
Open AccessDeep learning of topological phase transitions from entanglement aspects
The one-dimensional -wave superconductor proposed by Kitaev has long been a classic example for understanding topological phase transitions through various methods, such as examining the Berry phase, edge states of open chains, and, in particular, aspects from quantum entanglement of ground states. In order to understand the amount of information carried in the entanglement-related quantities, here we study topological phase transitions of the model with emphasis of using the deep learning approach. We feed different quantities, including Majorana correlation matrices (MCMs), entanglement spectra (ES) or entanglement eigenvectors (EE) originating from Block correlation matrices, into the deep neural networks for training, and investigate which one could be the most useful input format in this approach. We find that ES is information that is too compressed compared to MCM or EE. MCM and EE can provide us abundant information to recognize not only the topological phase transitions in the model but also phases of matter with different (1) gauges, which is not reachable by using ES only. - Open Access
Deep learning of topological phase transitions from entanglement aspects
The one-dimensional $p$-wave superconductor proposed by Kitaev has long been a classic example for understanding topological phase transitions through various methods, such as examining Berry phase, edge states of open chains and, in particular, aspects from quantum entanglement of ground states. In order to understand the amount of information carried in the entanglement-related quantities, here we study topological phase transitions of the model with emphasis of using the deep learning approach. We feed different quantities, including Majorana correlation matrices (MCMs), entanglement spectra (ES) or entanglement eigenvectors (EE) originated from Block correlation matrices (BCMs), into the deep neural networks for training, and investigate which one could be the most useful input format in this approach. We find that ES is indeed too compressed information compared to MCM or EE. MCM and EE can provide us abundant information to recognize not only the topological phase transitions in the model but also phases of matter with different $U$(1) gauges, which is not reachable by using ES only.
Open AccessDeep learning of topological phase transitions from entanglement aspects
The one-dimensional $p$-wave superconductor proposed by Kitaev has long been a classic example for understanding topological phase transitions through various methods, such as examining Berry phase, edge states of open chains and, in particular, aspects from quantum entanglement of ground states. In order to understand the amount of information carried in the entanglement-related quantities, here we study topological phase transitions of the model with emphasis of using the deep learning approach. We feed different quantities, including Majorana correlation matrices (MCMs), entanglement spectra (ES) or entanglement eigenvectors (EE) originated from Block correlation matrices (BCMs), into the deep neural networks for training, and investigate which one could be the most useful input format in this approach. We find that ES is indeed too compressed information compared to MCM or EE. MCM and EE can provide us abundant information to recognize not only the topological phase transitions in the model but also phases of matter with different $U$(1) gauges, which is not reachable by using ES only.