KUKeishu Utimula

Keishu Utimula

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Keishu Utimula has a total of 18 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Guideless Artificial Life Model for Reproduction, Development, and Interactions

    by Keishu Utimula

    Reproduction, development, and individual interactions are vital yet complex natural processes. Tierra (an ALife model proposed by Thomas Ray) and cellular automata, which can manage these aspects in a complex manner, are significantly limited in their ability to express morphology and behavior. In contrast, the virtual creatures proposed by Karl Sims have a considerably higher degree of freedom in terms of morphology and behavior. However, they also exhibit a limited capacity for processes like reproduction, development, and individual interactions. In addition, they employ genetic algorithms, which can result in a loss of biological diversity, as their implementation necessitates predefining a fitness function. Contrarily, the evolution of natural life is determined by mutation and natural selection, rather than by a human-defined fitness function. This study carefully extracts the characteristics of these models to propose a new Artificial Life model that can simulate reproduction, development, and individual interactions while exhibiting a high expressive power for morphology and behavior. The model is based on the concept of incorporating Tierra and cellular automata mechanisms into a cell that moves freely in 3-D space. In this model, no predefined fitness function or form that qualifies as a living creature exists. In other words, this approach can be rephrased as searching for persistent patterns, which is similar to the approach of Conway’s Game of Life. The primary objective of this study was to conduct a proof-of-concept demonstration to showcase the capabilities of this model. Guideless simulation by the proposed model using mutation and natural selection resulted in the formation of two types of creatures—dumbbell shaped and reticulated. These creatures exhibit intriguing features, exploiting the degrees of freedom inherent to the proposed model. Particularly noteworthy is their unique method of reproduction, which bears a striking resemblance to that of real organisms. These results reinforce the potential of this approach in modeling intricate processes observed in actual organisms and its ability to generate virtual creatures with intriguing ecologies.


  • Open Access

    Using reinforcement learning to autonomously identify sources of error for agents in group missions

    by Keishu Utimula, Ken-Taro Hayaschi, Trevor J. Bihl, Kenta Hongo, Ryo Maezono

    When agents swarm to execute a mission, some of them frequently exhibit sudden failure, as observed from the command base. It is generally difficult to determine whether a failure is caused by actuators (hypothesis, $h_a$) or sensors (hypothesis, $h_s$) by solely relying on the communication between the command base and concerning agent. However, by instigating collusion between the agents, the cause of failure can be identified; in other words, we expect to detect corresponding displacements for $h_a$ but not for $h_s$. In this study, we considered the question as to whether artificial intelligence can autonomously generate an action plan $\boldsymbol{g}$ to pinpoint the cause as aforedescribed. Because the expected response to $\boldsymbol{g}$ generally depends upon the adopted hypothesis [let the difference be denoted by $D(\boldsymbol{g})$], a formulation that uses $D\left(\boldsymbol{g}\right)$ to pinpoint the cause can be made. Although a $\boldsymbol{g}^*$ that maximizes $D(\boldsymbol{g})$ would be a suitable action plan for this task, such an optimization is difficult to achieve using the conventional gradient method, as $D(\boldsymbol{g})$ becomes nonzero in rare events such as collisions with other agents, and most swarm actions $\boldsymbol{g}$ give $D(\boldsymbol{g})=0$. In other words, throughout almost the entire space of $\boldsymbol{g}$, $D(\boldsymbol{g})$ has zero gradient, and the gradient method is not applicable. To overcome this problem, we formulated an action plan using Q-table reinforcement learning. Surprisingly, the optimal action plan generated via reinforcement learning presented a human-like solution to pinpoint the problem by colliding other agents with the failed agent. Using this simple prototype, we demonstrated the potential of applying Q-table reinforcement learning methods to plan autonomous actions to pinpoint the causes of failure.

  • Open Access

    Feature Space of XRD Patterns Constructed by an Autoencoder

    by Keishu Utimula, Masao Yano, Hiroyuki Kimoto, Kenta Hongo, Kousuke Nakano, Ryo Maezono

    X‐ray diffraction (XRD) is commonly used to analyze systematic variations occurring in compounds to tune their material properties. Machine learning can be used to extract such significant systematics among a series of observed peak patterns. The feature space concept, in the context of autoencoders, can be the platform for performing such extractions, where each peak pattern is projected into a space to extract the systematics. Herein, an autoencoder is trained to learn to detect the systematics driven by atomic substitutions within a single phase without structural transitions. The feature space constructed by the trained autoencoder classifies the substitution compositions of XRD patterns satisfactorily. The compositions interpolated in the feature space are in good agreement with those of an XRD pattern projected to a point. Subsequently, the autoencoder generates a virtual XRD pattern from an interpolated point in the feature space. When the feature space is effectively optimized by enough training data, the autoencoder predicts an XRD pattern with a concentration, which is difficult to be described using the possible resolution of the supercell method of ab initio calculations.


  • Open Access

    Shry: Application of Canonical Augmentation to the Atomic Substitution Problem

    by Genki Imam Prayogo, Andrea Tirelli, Keishu Utimula, Kenta Hongo, Ryo Maezono, Kousuke Nakano

    A common approach for studying a solid solution or disordered system within a periodic ab initio framework is to create a supercell in which certain amounts of target elements are substituted with other elements. The key to generating supercells is determining how to eliminate symmetry-equivalent structures from many substitution patterns. Although the total number of substitutions is on the order of trillions, only symmetry-inequivalent atomic substitution patterns need to be identified, and their number is far smaller than the total. Our developed Python software package, which is called Shry (Suite for High-throughput generation of models with atomic substitutions implemented by Python), allows the selection of only symmetry-inequivalent structures from the vast number of candidates based on the canonical augmentation algorithm. Shry is implemented in Python 3 and uses the CIF format as the standard for both reading and writing the reference and generated sets of substituted structures. Shry can be integrated into another Python program as a module or can be used as a stand-alone program. The implementation was verified through a comparison with other codes with the same functionality, based on the total numbers of symmetry-inequivalent structures, and also on the equivalencies of the output structures themselves. The provided crystal structure data used for the verification are expected to be useful for benchmarking other codes and also developing new algorithms in the future.


  • Open Access

    A quantum annealing approach to ionic diffusion in solids

    by Keishu Utimula, Tom Ichibha, Genki I. Prayogo, Kenta Hongo, Kousuke Nakano, Ryo Maezono

    We have developed a framework for using quantum annealing computation to evaluate a key quantity in ionic diffusion in solids, the correlation factor. Existing methods can only calculate the correlation factor analytically in the case of physically unrealistic models, making it difficult to relate microstructural information about diffusion path networks obtainable by current ab initio techniques to macroscopic quantities such as diffusion coefficients. We have mapped the problem into a quantum spin system described by the Ising Hamiltonian. By applying our framework in combination with ab initio technique, it is possible to understand how diffusion coefficients are controlled by temperatures, pressures, atomic substitutions, and other factors. We have calculated the correlation factor in a simple case with a known exact result by a variety of computational methods, including simulated quantum annealing on the spin models, the classical random walk, the matrix description, and quantum annealing on D-Wave with hybrid solver . This comparison shows that all the evaluations give consistent results with each other, but that many of the conventional approaches require infeasible computational costs. Quantum annealing is also currently infeasible because of the cost and scarcity of qubits, but we argue that when technological advances alter this situation, quantum annealing will easily outperform all existing methods.


  • Open Access

    Stochastic Estimations of the Total Number of Classes for a Clustering having Extremely Large Samples to be Included in the Clustering Engine

    by Keishu Utimula, Genki I. Prayogo, Kousuke Nakano, Kenta Hongo, Ryo Maezono

    Numerous reports have elucidated the classification of a large amount of data using various clustering techniques. However, an increase in data size hinders the applicability of these methods. Here, it is investigated how to deal with the exploding number of possibilities to be sorted into irreducible classes by using a clustering technique when its input capacity cannot accommodate the total number of possibilities. This can be exemplified by atomic substitutions in the supercell modeling of alloys. The number of possibilities is sometimes equal to trillions, which is extremely large to be accommodated in a cluster. Thus, it is not practically feasible to identify directly how many irreducible classes exist even though several techniques are available to perform the clustering. In this regard, a stochastic framework is developed to avoid the shortage limitations, providing a method to estimate the total number of irreducible classes (the order of classes), as a statistical estimate. The main conclusion is that the statistical variation of the number of classes, at each sampling trial, can serve as a promising measure to estimate the total number of irreducible classes. Characteristics of this approach is also discussed by comparing with the conventional one based on Polya's theorem.


  • Open Access

    Stochastic estimations of a total number of classes for the clusterings with too enormous samples to be accommodate into a clustering engine

    by Keishu Utimula, Genki I. Prayogo, Kousuke Nakano, Kenta Hongo, Ryo Maezono

    We considered the problem how to handle the exploding number of possibilities to be sorted into irreducible classes by using a clustering tool when its input capacity cannot accommodate the total number of the possibility. Concrete situations are explained taking examples of atomic substitutions in the supercell modeling of alloys. The number of the possibility sometimes amounts to $\sim$ trillion being too large to be accommodate. It is hence not practically feasible to identify how many irreducible classes exist by straightforward manner even though there are several tools available to perform the clustering. We have developed a stochastic framework to avoid the shortage of capacity, providing a method to estimate the total number of irreducible classes (the order of the classes) as a statistical estimate. A prominent conclusion derived here is that the statistical variation of the number of classes at each sampling trial is working as a promising measure to estimate the order.

  • Open Access

    Quantum annealing approach to Ionic Diffusion in Solid

    by Keishu Utimula, Tom Ichibha, Genki I. Prayogo, Kenta Hongo, Kousuke Nakano, Ryo Maezono

    We have developed a framework for using quantum annealing computation to evaluate a key quantity in ionic diffusion in solids, the correlation factor. Existing methods can only calculate the correlation factor analytically in the case of physically unrealistic models, making it difficult to relate microstructural information about diffusion path networks obtainable by current ${ab\ initio}$ techniques to macroscopic quantities such as diffusion coefficients. We have mapped the problem into a quantum spin system described by the Ising Hamiltonian. By applying our framework in combination with ab initio technique, it is possible to understand how diffusion coefficients are controlled by temperatures, pressures, atomic substitutions, and other factors.We have calculated the correlation factor in a simple case with a known exact result by a variety of computational methods, including simulated quantum annealing on the spin models, the classical random walk, the matrix description, and quantum annealing on D-Wave with hybrid solver. This comparison shows that all the evaluations give consistent results with each other, but that many of the conventional approaches require infeasible computational costs. Quantum annealing is also currently infeasible because of the cost and scarcity of Q-bits, but we argue that when technological advances alter this situation, quantum annealing will easily outperform all existing methods.

  • Open Access

    Machine‐Learning Clustering Technique Applied to Powder X‐Ray Diffraction Patterns to Distinguish Compositions of ThMn12‐Type Alloys

    by Keishu Utimula, Rutchapon Hunkao, Masao Yano, Hiroyuki Kimoto, Kenta Hongo, Shogo Kawaguchi, Sujin Suwanna, Ryo Maezono

    A clustering technique is applied using dynamic‐time‐wrapping (DTW) analysis to X‐ray diffraction (XRD) spectrum patterns in order to identify the microscopic structures of substituents introduced into the main phase of magnetic alloys. The clustering technique is found to perform well, identifying the concentrations of the substituents with success rates of ≈90%. This level of performance is attributed to the capability of DTW processing to filter out irrelevant information such as the peak intensities (due to the uncontrollability of diffraction conditions in polycrystalline samples) and the uniform shift of peak positions (due to the thermal expansion of lattices). The established framework is not limited to the system treated in this work, but is widely applicable to systems the properties of which are to be tuned by atomic substitutions within a phase. The framework has a broader potential to predict properties such as magnetic moments, optical spectra etc.) from observed XRD patterns, by predicting such properties evaluated from predicted microscopic local structure.


  • Open Access

    Feature space of XRD patterns constructed by auto-encorder

    by Keishu Utimula, Masao Yano, Hiroyuki Kimoto, Kenta Hongo, Kousuke Nakano, Ryo Maezono

    It would be a natural expectation that only major peaks, not all of them, would make an important contribution to the characterization of the XRD pattern. We developed a scheme that can identify which peaks are relavant to what extent by using auto-encoder technique to construct a feature space for the XRD peak patterns. Individual XRD patterns are projected onto a single point in the two-dimensional feature space constructed using the method. If the point is significantly shifted when a peak of interest is masked, then we can say the peak is relevant for the characterization represented by the point on the space. In this way, we can formulate the relevancy quantitatively. By using this scheme, we actually found such a peak with a significant peak intensity but low relevancy in the characterization of the structure. The peak is not easily explained by the physical viewpoint such as the higher-order peaks from the same plane index, being a heuristic finding by the power of machine-learning.