Yuta Suzuki
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Yuta Suzuki has a total of 31 co-authors. Here are the top co-authors by number of shared publications:
- Kanta Ono
Department of Applied Physics, The University of Osaka, Osaka, Japan
4 shared publications - Satoshi Yamane
Shimonoseki City University, Japan
3 shared publications - Tatsunori Taniai
OMRON SINIC X Corporation, Bunkyo, Tokyo 113-0033, Japan
3 shared publications - Yoshitaka Ushiku
Omron Sinic X Corporation, Tokyo, Japan
3 shared publications Kotaro Saito
Randeft Inc., Chiyoda, Japan
2 shared publications- Kotaro Saito
Laboratoire Léon Brillouin, CEA-CNRS, CEA/Saclay, 91191 Gif sur Yvette, France
2 shared publications
Latest Publications
- Open Access
- 17 June 2025
- Published by IOP Publishing in Machine Learning: Science and Technology
- Vol. 6 (3)
Bridging Text and Crystal Structures: Literature-driven Contrastive Learning for Materials Science
Understanding structure–property relationships is an essential yet challenging aspect of materials discovery and development. To facilitate this process, recent studies in materials informatics have sought latent embedding spaces of crystal structures to capture their similarities based on properties and functionalities. However, abstract feature-based embedding spaces are human-unfriendly and prevent intuitive and efficient exploration of the vast materials space. Here we introduce Contrastive Language--Structure Pre-training (CLaSP), a learning paradigm for constructing crossmodal embedding spaces between crystal structures and texts. CLaSP aims to achieve material embeddings that 1) capture property- and functionality-related similarities between crystal structures and 2) allow intuitive retrieval of materials via user-provided description texts as queries. To compensate for the lack of sufficient datasets linking crystal structures with textual descriptions, CLaSP leverages a dataset of over 400,000 published crystal structures and corresponding publication records, including paper titles and abstracts, for training. We demonstrate the effectiveness of CLaSP through text-based crystal structure screening and embedding space visualization.
- 17 June 2025
- Published by IOP Publishing in Machine Learning: Science and Technology
- Vol. 6 (3)
Open AccessBridging Text and Crystal Structures: Literature-driven Contrastive Learning for Materials Science
Understanding structure–property relationships is an essential yet challenging aspect of materials discovery and development. To facilitate this process, recent studies in materials informatics have sought latent embedding spaces of crystal structures to capture their similarities based on properties and functionalities. However, abstract feature-based embedding spaces are human-unfriendly and prevent intuitive and efficient exploration of the vast materials space. Here we introduce Contrastive Language--Structure Pre-training (CLaSP), a learning paradigm for constructing crossmodal embedding spaces between crystal structures and texts. CLaSP aims to achieve material embeddings that 1) capture property- and functionality-related similarities between crystal structures and 2) allow intuitive retrieval of materials via user-provided description texts as queries. To compensate for the lack of sufficient datasets linking crystal structures with textual descriptions, CLaSP leverages a dataset of over 400,000 published crystal structures and corresponding publication records, including paper titles and abstracts, for training. We demonstrate the effectiveness of CLaSP through text-based crystal structure screening and embedding space visualization.
- Open Access
- 12 December 2023
- Published by Springer Nature in Communications Materials
- Vol. 4 (1) , 1-12
Neural structure fields with application to crystal structure autoencoders
Representing crystal structures of materials to facilitate determining them via neural networks is crucial for enabling machine-learning applications involving crystal structure estimation. Among these applications, the inverse design of materials can contribute to explore materials with desired properties without relying on luck or serendipity. Here, we propose neural structure fields (NeSF) as an accurate and practical approach for representing crystal structures using neural networks. Inspired by the concepts of vector fields in physics and implicit neural representations in computer vision, the proposed NeSF considers a crystal structure as a continuous field rather than as a discrete set of atoms. Unlike existing grid-based discretized spatial representations, the NeSF overcomes the tradeoff between spatial resolution and computational complexity and can represent any crystal structure. We propose an autoencoder of crystal structures that can recover various crystal structures, such as those of perovskite structure materials and cuprate superconductors. Extensive quantitative results demonstrate the superior performance of the NeSF compared with the existing grid-based approach.
- 12 December 2023
- Published by Springer Nature in Communications Materials
- Vol. 4 (1) , 1-12
Open AccessNeural structure fields with application to crystal structure autoencoders
Representing crystal structures of materials to facilitate determining them via neural networks is crucial for enabling machine-learning applications involving crystal structure estimation. Among these applications, the inverse design of materials can contribute to explore materials with desired properties without relying on luck or serendipity. Here, we propose neural structure fields (NeSF) as an accurate and practical approach for representing crystal structures using neural networks. Inspired by the concepts of vector fields in physics and implicit neural representations in computer vision, the proposed NeSF considers a crystal structure as a continuous field rather than as a discrete set of atoms. Unlike existing grid-based discretized spatial representations, the NeSF overcomes the tradeoff between spatial resolution and computational complexity and can represent any crystal structure. We propose an autoencoder of crystal structures that can recover various crystal structures, such as those of perovskite structure materials and cuprate superconductors. Extensive quantitative results demonstrate the superior performance of the NeSF compared with the existing grid-based approach.
- Open Access
- 1 December 2022
- Published by IOP Publishing in Machine Learning: Science and Technology
- Vol. 3 (4) , 045034
Self-supervised learning of materials concepts from crystal structures via deep neural networks
Material development involves laborious processes to explore the vast materials space. The key to accelerating these processes is understanding the structure-functionality relationships of materials. Machine learning has enabled large-scale analysis of underlying relationships between materials via their vector representations, or embeddings. However, the learning of material embeddings spanning most known inorganic materials has remained largely unexplored due to the expert knowledge and efforts required to annotate large-scale materials data. Here we show that our self-supervised deep learning approach can successfully learn material embeddings from crystal structures of over 120 000 materials, without any annotations, to capture the structure-functionality relationships among materials. These embeddings revealed the profound similarity between materials, or ‘materials concepts’, such as cuprate superconductors and lithium-ion battery materials from the unannotated structural data. Consequently, our results enable us to both draw a large-scale map of the materials space, capturing various materials concepts, and measure the functionality-aware similarities between materials. Our findings will enable more strategic approaches to material development.
- 1 December 2022
- Published by IOP Publishing in Machine Learning: Science and Technology
- Vol. 3 (4) , 045034
Open AccessSelf-supervised learning of materials concepts from crystal structures via deep neural networks
Material development involves laborious processes to explore the vast materials space. The key to accelerating these processes is understanding the structure-functionality relationships of materials. Machine learning has enabled large-scale analysis of underlying relationships between materials via their vector representations, or embeddings. However, the learning of material embeddings spanning most known inorganic materials has remained largely unexplored due to the expert knowledge and efforts required to annotate large-scale materials data. Here we show that our self-supervised deep learning approach can successfully learn material embeddings from crystal structures of over 120 000 materials, without any annotations, to capture the structure-functionality relationships among materials. These embeddings revealed the profound similarity between materials, or ‘materials concepts’, such as cuprate superconductors and lithium-ion battery materials from the unannotated structural data. Consequently, our results enable us to both draw a large-scale map of the materials space, capturing various materials concepts, and measure the functionality-aware similarities between materials. Our findings will enable more strategic approaches to material development.
- 1 November 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 464-467
Single and Ensemble CNN Models with Out-Category Penalty in Cifar 10
In recent years, CNN have been used in many image recognition tasks. However, most of these CNN models learn only the features of the image, and do not learn the meta-information of the image. In this study, we proposed CNN models that can learn not only image features but also meta-information such as animals and vehicles by imposing an out-category penalty on the cifar10 dataset. As a result, our proposed model was found to be able to learn with meta-information and produce higher accuracy than existing CNN models.
- 1 November 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 464-467
Single and Ensemble CNN Models with Out-Category Penalty in Cifar 10
In recent years, CNN have been used in many image recognition tasks. However, most of these CNN models learn only the features of the image, and do not learn the meta-information of the image. In this study, we proposed CNN models that can learn not only image features but also meta-information such as animals and vehicles by imposing an out-category penalty on the cifar10 dataset. As a result, our proposed model was found to be able to learn with meta-information and produce higher accuracy than existing CNN models.
- 1 November 2020
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 125-129
Image Classification with Additional Non-decision Labels using Self-supervised learning and GAN
In recent years, image recognition has been improved by using machine learning to extract high accuracy. However, in most studies, it is assumed that the image input to a model has an answer. But then, when an unanswered image or an image of extremely low quality is entered into real space, the model is forced to answer and often makes a mistake. In this paper, we propose a method to discriminate between unscrutinized images using self-supervised learning and an image generation model using a subtask.
- 1 November 2020
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 125-129
Image Classification with Additional Non-decision Labels using Self-supervised learning and GAN
In recent years, image recognition has been improved by using machine learning to extract high accuracy. However, in most studies, it is assumed that the image input to a model has an answer. But then, when an unanswered image or an image of extremely low quality is entered into real space, the model is forced to answer and often makes a mistake. In this paper, we propose a method to discriminate between unscrutinized images using self-supervised learning and an image generation model using a subtask.
- 1 November 2020
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 483-485
An Image Classification Model that Learns MNIST Image Features and Numerical Information
In this study, we focused on the differences between image features and meta-information. For example, "0" and "9" are relatively similar in terms of image characteristics, but there is significant difference in terms of the numbers they actually mean. In contrast, "3" and "4" are relatively dissimilar in terms of image features, but the difference is small in terms of the values they actually mean. Therefore, we propose a method to learn not only the features of the image but also the actual meaning of the values. As a result, we were able to reduce the error.
- 1 November 2020
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 483-485
An Image Classification Model that Learns MNIST Image Features and Numerical Information
In this study, we focused on the differences between image features and meta-information. For example, "0" and "9" are relatively similar in terms of image characteristics, but there is significant difference in terms of the numbers they actually mean. In contrast, "3" and "4" are relatively dissimilar in terms of image features, but the difference is small in terms of the values they actually mean. Therefore, we propose a method to learn not only the features of the image but also the actual meaning of the values. As a result, we were able to reduce the error.
- Open Access
- 5 June 2020
- Published by Springer Nature in npj Computational Materials
- Vol. 6 (1) , 1-7
Automated crystal structure analysis based on blackbox optimisation
In the present study, we show that time-consuming manual tuning of parameters in the Rietveld method, one of the most frequently used crystal structure analysis methods in materials science, can be automated by considering the entire trial-and-error process as a blackbox optimisation problem. The automation is successfully achieved using Bayesian optimisation, which outperforms both a human expert and an expert-system type automation despite the absence of expertise. This approach stabilises the analysis quality by eliminating human-origin variance and bias, and can be applied to various analysis methods in other areas which also suffer from similar tiresome and unsystematic manual tuning of extrinsic parameters and human-origin variance and bias.
- 5 June 2020
- Published by Springer Nature in npj Computational Materials
- Vol. 6 (1) , 1-7
Open AccessAutomated crystal structure analysis based on blackbox optimisation
In the present study, we show that time-consuming manual tuning of parameters in the Rietveld method, one of the most frequently used crystal structure analysis methods in materials science, can be automated by considering the entire trial-and-error process as a blackbox optimisation problem. The automation is successfully achieved using Bayesian optimisation, which outperforms both a human expert and an expert-system type automation despite the absence of expertise. This approach stabilises the analysis quality by eliminating human-origin variance and bias, and can be applied to various analysis methods in other areas which also suffer from similar tiresome and unsystematic manual tuning of extrinsic parameters and human-origin variance and bias.
- Open Access
- 18 February 2019
- Published by AIP Publishing in Applied Physics Letters
- Vol. 114 (7) , 072404
Fabrication of L1-FeNi by pulsed-laser deposition
We demonstrated the fabrication of a rare-earth-free ferromagnetic L10-type Fe–Ni alloy (L10-FeNi) by pulsed laser deposition (PLD). We deposited Fe and Ni on Cu(001) by alternating monoatomic deposition via automatically stabilized laser ablation. We examined the structural properties, magnetic properties, and surface morphology of the alloy specimens as the growth temperature (Ts) was varied. We adequately confirmed the construction of the most prominent L10-FeNi phase at 300 °C, which is significantly higher than previously reported growth temperatures, indicating that PLD followed by thermal treatment promoted two-dimensional growth of the adsorbent. The formation process of L10-FeNi was investigated from the standpoint of surface thermodynamics, and the results suggest that the surface free energy of PLD and its highly instantaneous deposition process by PLD played key roles. Our findings are expected to lead to advanced methods for the fabrication of L10-FeNi.
- 18 February 2019
- Published by AIP Publishing in Applied Physics Letters
- Vol. 114 (7) , 072404
Open AccessFabrication of L1-FeNi by pulsed-laser deposition
We demonstrated the fabrication of a rare-earth-free ferromagnetic L10-type Fe–Ni alloy (L10-FeNi) by pulsed laser deposition (PLD). We deposited Fe and Ni on Cu(001) by alternating monoatomic deposition via automatically stabilized laser ablation. We examined the structural properties, magnetic properties, and surface morphology of the alloy specimens as the growth temperature (Ts) was varied. We adequately confirmed the construction of the most prominent L10-FeNi phase at 300 °C, which is significantly higher than previously reported growth temperatures, indicating that PLD followed by thermal treatment promoted two-dimensional growth of the adsorbent. The formation process of L10-FeNi was investigated from the standpoint of surface thermodynamics, and the results suggest that the surface free energy of PLD and its highly instantaneous deposition process by PLD played key roles. Our findings are expected to lead to advanced methods for the fabrication of L10-FeNi.
- 14 May 2004
- Published by Wiley in Journal of Agronomy and Crop Science
- Vol. 190 (3) , 218-221
Differences in Starch Characteristics of Rice Strains Having Different Sensitivities to Maturation Temperatures
The amylose synthesis in rice with Wxb is affected by environmental temperature during seed development. We previously isolated a mutant candidate (coi) in which amylose synthesis is insensitive to a cool temperature. Here, the effects of maturation temperature (21 °C vs. 28 °C) on the starch characteristics of coi and its original strain 76‐3/T65 were studied. Amylose contents and gelatinization properties of 76‐3/T65 were affected by maturation temperatures, while coi was less affected. The amylopectin chain length distributions were almost the same between the two strains maturated at both temperatures. Gelatinization onset (To), peak (Tp), conclusion temperatures (Tc) and enthalpies (ΔH) decreased with environmental temperatures, and Tp, Tc and ΔH of coi at 21 °C were significantly higher than those of 76‐3/T65. Thus, the coi mutant, less affected by maturation temperature than that of 76‐3/T65, may be a useful strain in understanding biochemical and genetic starch biosynthesis response to slight changes in temperature.
- 14 May 2004
- Published by Wiley in Journal of Agronomy and Crop Science
- Vol. 190 (3) , 218-221
Differences in Starch Characteristics of Rice Strains Having Different Sensitivities to Maturation Temperatures
The amylose synthesis in rice with Wxb is affected by environmental temperature during seed development. We previously isolated a mutant candidate (coi) in which amylose synthesis is insensitive to a cool temperature. Here, the effects of maturation temperature (21 °C vs. 28 °C) on the starch characteristics of coi and its original strain 76‐3/T65 were studied. Amylose contents and gelatinization properties of 76‐3/T65 were affected by maturation temperatures, while coi was less affected. The amylopectin chain length distributions were almost the same between the two strains maturated at both temperatures. Gelatinization onset (To), peak (Tp), conclusion temperatures (Tc) and enthalpies (ΔH) decreased with environmental temperatures, and Tp, Tc and ΔH of coi at 21 °C were significantly higher than those of 76‐3/T65. Thus, the coi mutant, less affected by maturation temperature than that of 76‐3/T65, may be a useful strain in understanding biochemical and genetic starch biosynthesis response to slight changes in temperature.
- 1 December 1995
- Published by Elsevier in Thermochimica Acta
- Vol. 267, 231-237
Thermal degradation of aromatic polyformals derived from bisphenol A and bisphenol AF
- 1 December 1995
- Published by Elsevier in Thermochimica Acta
- Vol. 267, 231-237
Thermal degradation of aromatic polyformals derived from bisphenol A and bisphenol AF