Category: Machine learning

2DMAT is a framework for applying a search algorithm to a direct problem solver to find the optimal solution. As the standard direct problem solver, the experimental data analysis software for two-dimensional material structure analysis is prepared. The direct problem solver gives the deviation between the experimental data and the calculated data obtained under the given parameters such as atomic positions as a loss function used in the inverse problem. The optimal parameters are estimated by minimizing the loss function using a search algorithm. For further use, the original direct problem solver or the search algorithm can be defined by users. In the version 2.0, for solving a direct problem, 2DMAT offers the wrapper of the solver for the total-reflection high-energy positron diffraction (TRHEPD) experiment, sxrd, and leed. As algorithms, it offers the Nelder-Mead method, the grid search method, the Bayesian optimization method, the replica exchange Monte Carlo method, and the population annealing Monte Carlo method. This repository provides the sample data including input and output data for 2DMAT.

Moller is a program to support high-throughput computations, provided as a part of the HTP-tools package. The use of machine learning for predicting material properties and designing substances, known as materials informatics, has gained considerable attention in recent years. The accuracy of machine learning depends heavily on the preparation of appropriate training data. Therefore, the development of tools and environments for the rapid generation of training data is expected to contribute significantly to the advancement of research in materials informatics. Moller generates batch job scripts for supercomputers and clusters, allowing parallel execution of programs under a series of computational conditions, such as parameter parallelism. This repository provides sample inputs and outputs that illustrate the use of Moller.

Single-shot imaging with femtosecond X-ray lasers is a powerful measurement technique that can achieve both high spatial and temporal resolution. However, its accuracy has been severely limited by the difficulty of applying conventional background processing. In this data repository, sample codes, training dataset and learned model for deep learning method which was used to validate noise reduction techniques are provided.

PHYSBO (optimization tools for PHYSics based on Bayesian Optimization) is a Python library for fast and scalable Bayesian optimization. It has been developed mainly for application in the basic sciences such as physics and materials science. Bayesian optimization is used to select an appropriate input for experiments/simulations from candidate inputs listed in advance in order to obtain better output values with the help of machine learning prediction. PHYSBO can be used to find better solutions for both single and multi-objective optimization problems. At each cycle in the Bayesian optimization, a single proposal or multiple proposals can be obtained for the next experiments/simulations. These proposals can be obtained interactively for use in experiments. PHYSBO is available at https://github.com/issp-center-dev/PHYSBO. This repository provides the sample data including input and output data for PHYSBO.

abICS is a software framework for training a machine learning model to reproduce first-principles energies and then using the model to perform configurational sampling in disordered systems. Specific emphasis is placed on multi-component solid state systems such as metal and oxide alloys. The current version of abics can use neural network models implemented in aenet to be used as the machine learning model. As of this moment, abICS can also generate Quantum Espresso, VASP, and OpenMX input files for obtaining the reference training data for the machine learning model. This repository provides the sample data including input and output data for abICS.

Training data for neural network potential of amorphous GeO2, as well as resulting molecular dynamics data and structure characterization

Molecular dynamics trajectory for proton diffusion in heavily Sc-doped BaSnO3, the force field used for the simulation, and training data for the machine-learning force field.

Analytic continuation (AC) from the imaginary-time Green’s function to the spectral function is a crucial process for numerical studies of the dynamical properties of quantum many-body systems. This process, however, is an ill-posed problem; that is, the obtained spectrum is unstable against the noise of the Green’s function. Though several numerical methods have been developed, each of them has its own advantages and disadvantages. The sparse modeling (SpM) AC method, for example, is robust against the noise of the Green’s function but suffers from unphysical oscillations in the low energy region. We propose a new method that combines the SpM AC with the Pade approximation. This combination, called SpM-Pade, inherits robustness against noise from SpM and low-energy accuracy from Pade, compensating for the disadvantages of each. We demonstrate that the SpMPad´e method yields low-variance and low-biased results with almost the same computational cost as that of the SpM method.