Nguyen Duong Nguyen
@jaist.ac.jp
Knowledge Science school
Japan Advanced Institute of Science and Technology
RESEARCH, TEACHING, or OTHER INTERESTS
Materials Science, Computational Theory and Mathematics
Scopus Publications
Scopus Publications
Tien-Sinh Vu, Minh-Quyet Ha, Duong-Nguyen Nguyen, Viet-Cuong Nguyen, Yukihiro Abe, Truyen Tran, Huan Tran, Hiori Kino, Takashi Miyake, Koji Tsuda,et al.
Springer Science and Business Media LLC
AbstractDeep learning (DL) models currently employed in materials research exhibit certain limitations in delivering meaningful information for interpreting predictions and comprehending the relationships between structure and material properties. To address these limitations, we propose an interpretable DL architecture that incorporates the attention mechanism to predict material properties and gain insights into their structure–property relationships. The proposed architecture is evaluated using two well-known datasets (the QM9 and the Materials Project datasets), and three in-house-developed computational materials datasets. Train–test–split validations confirm that the models derived using the proposed DL architecture exhibit strong predictive capabilities, which are comparable to those of current state-of-the-art models. Furthermore, comparative validations, based on first-principles calculations, indicate that the degree of attention of the atoms’ local structures to the representation of the material structure is critical when interpreting structure–property relationships with respect to physical properties. These properties encompass molecular orbital energies and the formation energies of crystals. The proposed architecture shows great potential in accelerating material design by predicting material properties and explicitly identifying crucial features within the corresponding structures.
Duong-Nguyen Nguyen and Hieu-Chi Dam
AIP Publishing
We investigate the correlation between geometrical information, stability, and magnetization of SmFe12-based structures using machine learning-aided genetic algorithm structure generation and first-principle calculation. In parallel with structure generation inherited using the USPEX program, a pool of structures is created for every population using the sub-symmetry perturbation method. A framework using embedded orbital field matrix representation as structure fingerprint and Gaussian process as a predictor has been applied to ranking the most potential stability structures. As a result, the original structure SmFe12 with the well-known tetragonal I4/mmm symmetry is investigated with a parabolic dependence between formation energy and its magnetization by continuous distortions of the unit-cell lattice parameter and individual sites. Notably, a SmFe12 structure with I4/mmm symmetry is found with 7.5% increasing magnetization while keeping the similar formation energy with the most stable structures in this family. With SmFe11CoN family, structures with N interstitial position in the center of Sm and Fe octahedron show outperform all other structures in both ability of stabilization and remaining high magnetization of the original structure. Finally, further investigation using metric learning embedding space brings valuable insight into the correlation between geometrical arrangement, stability, and magnetization of this structure family.
Minh-Quyet Ha, Duong-Nguyen Nguyen, Viet-Cuong Nguyen, Hiori Kino, Yasunobu Ando, Takashi Miyake, Thierry Denœux, Van-Nam Huynh, and Hieu-Chi Dam
AIP Publishing
Measuring the similarity between materials is essential for estimating their properties and revealing the associated physical mechanisms. However, current methods for measuring the similarity between materials rely on theoretically derived descriptors and parameters fitted from experimental or computational data, which are often insufficient and biased. Furthermore, outliers and data generated by multiple mechanisms are usually included in the dataset, making the data-driven approach challenging and mathematically complicated. To overcome such issues, we apply the Dempster–Shafer theory to develop an evidential regression-based similarity measurement (eRSM) method, which can rationally transform data into evidence. It then combines such evidence to conclude the similarities between materials, considering their physical properties. To evaluate the eRSM, we used two material datasets, including 3d transition metal–4f rare-earth binary and quaternary high-entropy alloys with target properties, Curie temperature, and magnetization. Based on the information obtained on the similarities between the materials, a clustering technique is applied to learn the cluster structures of the materials that facilitate the interpretation of the mechanism. The unsupervised learning experiments demonstrate that the obtained similarities are applicable to detect anomalies and appropriately identify groups of materials whose properties correlate differently with their compositions. Furthermore, significant improvements in the accuracies of the predictions for the Curie temperature and magnetization of the quaternary alloys are obtained by introducing the similarities, with the reduction in mean absolute errors of 36% and 18%, respectively. The results show that the eRSM can adequately measure the similarities and dissimilarities between materials in these datasets with respect to mechanisms of the target properties.
Duong Dai Nguyen, Yuchen Deng, and Toshi-kazu Suzuki
IOP Publishing
Abstract We have systematically investigated low-frequency noise (LFN) in AlTiO/AlGaN/GaN metal-insulator-semiconductor field-effect transistors (FETs) with non-gate-recessed or partially-gate-recessed structures, where gate insulators using AlTiO, an alloy of Al2O3 and TiO2, are obtained by atomic layer deposition. For drain current LFN, we find pure 1 / f spectra for the well-above-threshold regime, and superposition of 1 / f and Lorentzian spectra near the threshold voltage. The Hooge parameters are evaluated from the 1 / f contribution and found to be independent of the AlTiO thickness. However, the remaining AlGaN thickness strongly affects the Hooge parameter near the threshold voltage; in the low channel electron concentration regime of the partially-gate-recessed FETs, a smaller remaining AlGaN thickness gives a larger Hooge parameter proportional to the inverse of the electron concentration, indicating that channel electron number fluctuation dominates the Hooge parameter. We consider that the channel electron number fluctuation is caused by electron traps introduced by the recess etching process in the remaining AlGaN. On the other hand, the Lorentzian spectra give specific time constants almost independent of the AlTiO thickness and the remaining AlGaN thickness, corresponding to trap depths of 0.6–0.8 eV. This can be attributed to traps in AlTiO near the AlTiO/AlGaN interface.
Duong-Nguyen Nguyen, Hiori Kino, Takashi Miyake, and Hieu-Chi Dam
Springer Science and Business Media LLC
Abstract In this article, we propose a query-and-learn active learning approach combined with first-principles calculations to rapidly search for potentially stable crystal structure via elemental substitution, to clarify their stabilization mechanism, and integrate this approach to SmFe$$_{12}$$ 12 -based compounds with ThMn$$_{12}$$ 12 structure, which exhibits prominent magnetic properties. The proposed method aims to (1) accurately estimate formation energies with limited first-principles calculation data, (2) visually monitor the progress of the structure search process, (3) extract correlations between structures and formation energies, and (4) recommend the most beneficial candidates of SmFe$$_{12}$$ 12 -substituted structures for the subsequent first-principles calculations. The structures of SmFe$$_{12-\\upalpha -\\upbeta }\\mathsf {X}_{\\upalpha }\\mathsf {Y}_{\\upbeta }$$ 12 - α - β X α Y β before optimization are prepared by substituting $$\\mathsf {X}, \\mathsf {Y}$$ X , Y elements—Mo, Zn, Co, Cu, Ti, Al, Ga—in the region of $$\\upalpha +\\upbeta <4$$ α + β < 4 into iron sites of the original SmFe$$_{12}$$ 12 structures. Using the optimized structures and formation energies obtained from the first-principles calculations after each active learning cycle, we construct an embedded two-dimensional space to rationally visualize the set of all the calculated and not-yet-calculated structures for monitoring the progress of the search. Our machine learning model with an embedding representation attained a prediction error for the formation energy of $$1.25\\times 10^{-2}$$ 1.25 × 10 - 2 (eV/atom) and required only one-sixth of the training data compared to other learning methods. Moreover, the time required to recall most potentially stable structures was nearly four times faster than the random search. The formation energy landscape visualized using the embedding representation revealed that the substitutions of Al and Ga have the highest potential to stabilize the SmFe$$_{12}$$ 12 structure. In particular, SmFe$$_{9}$$ 9 [Al/Ga]$$_{2}$$ 2 Ti showed the highest stability among the investigated structures. Finally, by quantitatively measuring the change in the structures before and after optimization using OFM descriptors, the correlations between the coordination number of substitution sites and the resulting formation energy are revealed. The negative-formation-energy-family SmFe$$_{12-\\upalpha -\\upbeta }$$ 12 - α - β [Al/Ga]$$_{\\upalpha }\\mathsf {Y}_{\\upbeta }$$ α Y β structures show a common trend of increasing coordination number at substituted sites, whereas structures with positive formation energy show a corresponding decreasing trend. Impact statement Seeking the next generation of high-performance magnets is a crucial demand for replacing the widely accepted Nd-Fe-B magnets developed in the middle 80s. The iron-rich compounds with the original tetragonal ThMn12 structure appear as the most potential candidates except for the hard synthesizing it in nature due to its high energy of formation. Stabilization for this material system is expected by substituting new elements, but the vast number of possible structures makes the exploration difficult even for theoretical calculations. This article proposes an integration of first-principles calculations and explainable active learning to efficiently explore the crystal structure space of this material system. In particular, the explored crystal structure space can be rationally visualized, on which the relationship between substitution elements, substitution sites, and crystal structure stabilization can be intuitively interpreted.
Hirosuke Matsui, Nozomu Ishiguro, Yuanyuan Tan, Naoyuki Maejima, Yuta Muramoto, Tomoya Uruga, Kotaro Higashi, Duong‐Nguyen Nguyen, Hieu‐Chi Dam, Gabor Samjeské,et al.
Wiley
AbstractIn practical applications of polymer electrolyte fuel cells (PEFC), the wide variations of local structures and environment of electrocatalysts bring about complex reaction behaviours inside a membrane electrode assembly (MEA). The variations of local coordination structure and redox response of a Pt/C cathode catalyst in an MEA before and after a typical accelerated degradation test (ADT) were three‐dimensionally visualized by operando Pt LIII‐edge computed‐tomography X‐ray absorption fine structure (CT‐XAFS) imaging under PEFC operating conditions for the first time. The set of operando CT‐XANES and CT‐EXAFS analyses visualized changes in the local coordination structure and redox reactivity of the Pt catalyst by a typical ADT process.
Duong Dai Nguyen, Takehiro Isoda, Yuchen Deng, and Toshi-kazu Suzuki
AIP Publishing
We report normally-off operations in partially-gate-recessed AlxTiyO(AlTiO)/AlGaN/GaN metal-insulator-semiconductor (MIS) field-effect transistors (FETs), where aluminum titanium oxide AlTiO, an alloy of Al2O3 and TiO2, is employed as a gate insulator. Since AlTiO is useful for interface charge engineering owing to a trend that the AlTiO/AlGaN interface fixed charge is suppressed in comparison with Al2O3, we investigated combining the interface charge engineering with a partial gate recess method for AlTiO/AlGaN/GaN MIS-FETs. For AlTiO with a composition of x/(x+y)=0.73, a suppressed positive interface fixed charge at the AlTiO/recessed-AlGaN interface leads to a positive slope in the relation between the threshold voltage and the AlTiO insulator thickness. As a result, we successfully obtained normally-off operations in partially-gate-recessed AlTiO/AlGaN/GaN MIS-FETs with favorable performances, such as a threshold voltage of 1.7 V, an on-resistance of 9.5Ωmm, an output current of 450 mA/mm, a low sub-threshold swing of 65 mV/decade, and a rather high electron mobility of 730cm2/Vs. The results show that the interface charge engineering in combination with partial gate recess is effective for the GaN-based normally-off device technology.
Minh-Quyet Ha, Duong-Nguyen Nguyen, Viet-Cuong Nguyen, Takahiro Nagata, Toyohiro Chikyow, Hiori Kino, Takashi Miyake, Thierry Denœux, Van-Nam Huynh, and Hieu-Chi Dam
Springer Science and Business Media LLC
AbstractExisting data-driven approaches for exploring high-entropy alloys (HEAs) face three challenges: numerous element-combination candidates, designing appropriate descriptors, and limited and biased existing data. To overcome these issues, here we show the development of an evidence-based material recommender system (ERS) that adopts Dempster–Shafer theory, a general framework for reasoning with uncertainty. Herein, without using material descriptors, we model, collect and combine pieces of evidence from data about the HEA phase existence of alloys. To evaluate the ERS, we compared its HEA-recommendation capability with those of matrix-factorization- and supervised-learning-based recommender systems on four widely known datasets of up-to-five-component alloys. The k-fold cross-validation on the datasets suggests that the ERS outperforms all competitors. Furthermore, the ERS shows good extrapolation capabilities in recommending quaternary and quinary HEAs. We experimentally validated the most strongly recommended Fe–Co-based magnetic HEA (namely, FeCoMnNi) and confirmed that its thin film shows a body-centered cubic structure.
Duy Cong Le, Duong Dai Nguyen, Savanna Lloyd, Toshi-kazu Suzuki, and Hideyuki Murata
Royal Society of Chemistry (RSC)
The initial luminescence loss of fluorescent OLEDs utilizing triplet–triplet annihilation is not only caused by quenching of singlet exciton with neutral quenchers but also by that of triplet excitons with positively charged quenchers.
Tien-Lam Pham, Duong-Nguyen Nguyen, Minh-Quyet Ha, Hiori Kino, Takashi Miyake, and Hieu-Chi Dam
International Union of Crystallography (IUCr)
New Nd–Fe–B crystal structures can be formed via the elemental substitution of LA–T–X host structures, including lanthanides (LA), transition metals (T) and light elements, X = B, C, N and O. The 5967 samples of ternary LA–T–X materials that are collected are then used as the host structures. For each host crystal structure, a substituted crystal structure is created by substituting all lanthanide sites with Nd, all transition metal sites with Fe and all light-element sites with B. High-throughput first-principles calculations are applied to evaluate the phase stability of the newly created crystal structures, and 20 of them are found to be potentially formable. A data-driven approach based on supervised and unsupervised learning techniques is applied to estimate the stability and analyze the structure–stability relationship of the newly created Nd–Fe–B crystal structures. For predicting the stability for the newly created Nd–Fe–B structures, three supervised learning models: kernel ridge regression, logistic classification and decision tree model, are learned from the LA–T–X host crystal structures; the models achieved maximum accuracy and recall scores of 70.4 and 68.7%, respectively. On the other hand, our proposed unsupervised learning model based on the integration of descriptor-relevance analysis and a Gaussian mixture model achieved an accuracy and recall score of 72.9 and 82.1%, respectively, which are significantly better than those of the supervised models. While capturing and interpreting the structure–stability relationship of the Nd–Fe–B crystal structures, the unsupervised learning model indicates that the average atomic coordination number and coordination number of the Fe sites are the most important factors in determining the phase stability of the new substituted Nd–Fe–B crystal structures.
Duong-Nguyen Nguyen, Duc-Anh Dao, Takashi Miyake, and Hieu-Chi Dam
AIP Publishing
In this study, we investigate the structure–stability relationship of hypothetical Nd–Fe–B crystal structures using descriptor-relevance analysis and the t-SNE dimensionality reduction method. 149 hypothetical Nd–Fe–B crystal structures are generated from 5967 LA–T–X host structures in the Open Quantum Materials Database by using the elemental substitution method, with LA denoting lanthanides, T denoting transition metals, and X denoting light elements such as B, C, N, and O. By borrowing the skeletal structure of each of the host materials, a hypothetical crystal structure is created by substituting all lanthanide sites with Nd, all transition metal sites with Fe, and all light element sites with B. High-throughput first-principle calculations are applied to evaluate the phase stability of these structures. Twenty of them are found to be potentially formable. As the first investigative result, the descriptor-relevance analysis on the orbital field matrix (OFM) materials’ descriptor reveals the average atomic coordination number as the essential factor in determining the structure stability of these substituted Nd–Fe–B crystal structures. 19 among 20 hypothetical structures that are found potentially formable have an average coordination number larger than 6.5. By applying the t-SNE dimensionality reduction method, all the local structures represented by the OFM descriptors are integrated into a visible space to study the detailed correlation between their characteristics and the stability of the crystal structure to which they belong. We discover that unstable substituted structures frequently carry Nd and Fe local structures with two prominent points: low average coordination numbers and fully occupied B neighboring atoms. Moreover, there are only three popular forms of B local structures appearing on all potentially formable substituted structures: cage networks, planar networks, and interstitial sites. The discovered relationships are promising to speed up the screening process for the new formable crystal structures.
Duong Dai Nguyen and Toshi-kazu Suzuki
AIP Publishing
Toward interface charge engineering in AlTiO/AlGaN/GaN metal-insulator-semiconductor (MIS) devices, we systematically investigated insulator-semiconductor interface fixed charges depending on the composition of the AlTiO gate insulator obtained by atomic layer deposition. By evaluating the positive interface fixed charge density from the insulator-thickness dependence of the threshold voltages of the MIS devices, we found a trend that the interface fixed charge density decreases with the decrease in the Al composition ratio, i.e., increase in the Ti composition ratio, which leads to shallow threshold voltages. This trend can be attributed to the large bonding energy of O-Ti in comparison with that of O-Al and to consequent possible suppression of interface oxygen donors. For an AlTiO gate insulator with an intermediate composition, the MIS field-effect transistors exhibit favorable device characteristics with high linearity of transconductance. These results indicate a possibility of interface charge engineering using AlTiO, in addition to energy gap engineering and dielectric constant engineering.
Duong-Nguyen Nguyen, Tien-Lam Pham, Viet-Cuong Nguyen, Hiori Kino, Takashi Miyake, and Hieu-Chi Dam
IOP Publishing
In the original paper [ 1 we introduced a voting method for measuring dissimilarity materials with respect to a given target property. The technique is based on an ensemble method with Kernel ridge regression as the predicting model; multiple random subset sampling of the materials is done to generate prediction models and the corresponding contributions of the reference training materials in detail. The reference training materials contributed to the prediction model that accurately predicts the physical property value of a speci fi c material, are considered to be similar to that material, or vice versa.
Makoto Hirose, Nozomu Ishiguro, Kei Shimomura, Duong-Nguyen Nguyen, Hirosuke Matsui, Hieu Chi Dam, Mizuki Tada, and Yukio Takahashi
Springer Science and Business Media LLC
AbstractOxygen storage and release with oxygen diffusion in the bulk of the cerium–zirconium solid solution oxide Ce2Zr2Ox (x = 7–8), which possesses an atomically ordered arrangement of cerium and zirconium atoms, is the key to three-way exhaust catalysis. Oxygen storage proceeds via heterogeneous oxygen diffusion into the vacant sites of Ce2Zr2O7 particles, but the heterogeneous oxygen diffusion track is erased after oxygen storage in the Ce2Zr2Ox bulk. Here we show three-dimensional hard X-ray spectro-ptychography to clearly visualize the three-dimensional cerium valence map in Ce2Zr2Ox particles, and unsupervised learning reveals the concealed oxygen-diffusion-driven three-dimensional nanoscale cerium oxidation behavior and tracking areas inside individual mixed-oxide particles during the oxygen storage process. The described approach may permit the nanoscale chemical imaging of reaction tracking areas in solid materials.
Duong-Nguyen Nguyen, Tien-Lam Pham, Viet-Cuong Nguyen, Anh-Tuan Nguyen, Hiori Kino, Takashi Miyake, and Hieu-Chi Dam
IOP Publishing
Abstract The Curie temperature (T C) of RT binary compounds consisting of 3d transition-metal (T ) and 4f rare-earth elements (R) is analyzed systematically by a developed machine learning technique called kernel regression-based model evaluation. Twenty-one descriptive variables were designed assuming completely obtained information of the T C. Multiple kernel regression analyses with different kernel types: cosine, linear, Gaussian, polynomial, and Laplacian kernels were implemented and examined. All possible descriptive variable combinations were generated to construct the corresponding prediction models. As a result, by appropriate combinations between descriptive variable sets and kernel formulations, we demonstrate that a number of kernel regression models can accurately reproduce the T C of the RT compounds. The relevance of descriptive variables for predicting T C are systematically investigated. The results indicate that the rare-earth concentration is the most relevant variable in the T C phenomenon. We demonstrate that the regression-based model selection technique can be applied to learn the relationship between the descriptive variables and the actuation mechanism of the corresponding physical phenomenon, i.e., T C in the present case.
Yuanyuan Tan, Hirosuke Matsui, Nozomu Ishiguro, Tomoya Uruga, Duong-Nguyen Nguyen, Oki Sekizawa, Tomohiro Sakata, Naoyuki Maejima, Kotaro Higashi, Hieu Chi Dam,et al.
American Chemical Society (ACS)
Catalyst degradation at the cathode of a membrane electrode assembly (MEA) remains a critical issue for practical polymer electrolyte fuel cell (PEFC) operation, but such wet systems impede detaile...
Duong-Nguyen Nguyen, Tien-Lam Pham, Viet-Cuong Nguyen, Hiori Kino, Takashi Miyake, and Hieu-Chi Dam
IOP Publishing
Abstract We propose a data-driven method to extract dissimilarity between materials, with respect to a given target physical property. The technique is based on an ensemble method with Kernel ridge regression as the predicting model; multiple random subset sampling of the materials is done to generate prediction models and the corresponding contributions of the reference training materials in detail. The distribution of the predicted values for each material can be approximated by a Gaussian mixture models. The reference training materials contributed to the prediction model that accurately predicts the physical property value of a specific material, are considered to be similar to that material, or vice versa. Evaluations using synthesized data demonstrate that the proposed method can effectively measure the dissimilarity between data instances. An application of the analysis method on the data of Curie temperature ( T C ) of binary 3d transition metal- 4f rare-earth binary alloys also reveals meaningful results on the relations between the materials. The proposed method can be considered as a potential tool for obtaining a deeper understanding of the structure of data, with respect to a target property, in particular.
Duong Nam Nguyen, Duong Nguyen Nguyen, and Mai Khanh Pham
SciCell
In this paper, the influence of rare earth (RE) on the microstructure and mechanical properties of austenitic high manganese steel (HMnS) Mn15Cr2V were investigated. The results showed that the microstructure, hardness and impact strength of RE modification sample is finer and better than non-modified sample. Under the effect of impact load, the hardness and the depth of the work-hardening layer of the modified steel was higher than that of the non-modified steel, thereby, the value of microhardness in the surface of the modified sample was 420 HV while it was only 395 HV in the non-modified sample. The value of the impact strength of the modified sample was up to 132 J/cm2 compared to the non-modified sample is only 115 J/cm2. Moreover, after impact load, the austenite nanoparticles had been found out on the surface of this steel, this is the cause of the increasing of mechanical properties in this steel.
Tien-Lam Pham, Nguyen-Duong Nguyen, Van-Doan Nguyen, Hiori Kino, Takashi Miyake, and Hieu-Chi Dam
AIP Publishing
We have developed a descriptor named Orbital Field Matrix (OFM) for representing material structures in datasets of multi-element materials. The descriptor is based on the information regarding atomic valence shell electrons and their coordination. In this work, we develop an extension of OFM called OFM1. We have shown that these descriptors are highly applicable in predicting the physical properties of materials and in providing insights on the materials space by mapping into a low embedded dimensional space. Our experiments with transition metal/lanthanide metal alloys show that the local magnetic moments and formation energies can be accurately reproduced using simple nearest-neighbor regression, thus confirming the relevance of our descriptors. Using kernel ridge regressions, we could accurately reproduce formation energies and local magnetic moments calculated based on first-principles, with mean absolute errors of 0.03 μB and 0.10 eV/atom, respectively. We show that meaningful low-dimensional representations can be extracted from the original descriptor using descriptive learning algorithms. Intuitive prehension on the materials space, qualitative evaluation on the similarities in local structures or crystalline materials, and inference in the designing of new materials by element substitution can be performed effectively based on these low-dimensional representations.
Son Phuong Le, Duong Dai Nguyen, and Toshi-kazu Suzuki
AIP Publishing
We have investigated insulator-semiconductor interface fixed charges in AlGaN/GaN metal-insulator-semiconductor (MIS) devices with Al2O3 or AlTiO (an alloy of Al2O3 and TiO2) gate dielectrics obtained by atomic layer deposition on AlGaN. Analyzing insulator-thickness dependences of threshold voltages for the MIS devices, we evaluated positive interface fixed charges, whose density at the AlTiO/AlGaN interface is significantly lower than that at the Al2O3/AlGaN interface. This and a higher dielectric constant of AlTiO lead to rather shallower threshold voltages for the AlTiO gate dielectric than for Al2O3. The lower interface fixed charge density also leads to the fact that the two-dimensional electron concentration is a decreasing function of the insulator thickness for AlTiO, whereas being an increasing function for Al2O3. Moreover, we discuss the relationship between the interface fixed charges and interface states. From the conductance method, it is shown that the interface state densities are very similar at the Al2O3/AlGaN and AlTiO/AlGaN interfaces. Therefore, we consider that the lower AlTiO/AlGaN interface fixed charge density is not owing to electrons trapped at deep interface states compensating the positive fixed charges and can be attributed to a lower density of oxygen-related interface donors.
Duong-Nguyen Nguyen, Tien-Lam Pham, Viet-Cuong Nguyen, Tuan-Dung Ho, Truyen Tran, Keisuke Takahashi, and Hieu-Chi Dam
International Union of Crystallography (IUCr)
A method has been developed to measure the similarity between materials, focusing on specific physical properties. The information obtained can be utilized to understand the underlying mechanisms and support the prediction of the physical properties of materials. The method consists of three steps: variable evaluation based on nonlinear regression, regression-based clustering, and similarity measurement with a committee machine constructed from the clustering results. Three data sets of well characterized crystalline materials represented by critical atomic predicting variables are used as test beds. Herein, the focus is on the formation energy, lattice parameter and Curie temperature of the examined materials. Based on the information obtained on the similarities between the materials, a hierarchical clustering technique is applied to learn the cluster structures of the materials that facilitate interpretation of the mechanism, and an improvement in the regression models is introduced to predict the physical properties of the materials. The experiments show that rational and meaningful group structures can be obtained and that the prediction accuracy of the materials' physical properties can be significantly increased, confirming the rationality of the proposed similarity measure.