-
Wavelet Scattering Networks for Atomistic Systems with Extrapolation of Material Properties
Authors:
Paul Sinz,
Michael W. Swift,
Xavier Brumwell,
Jialin Liu,
Kwang ** Kim,
Yue Qi,
Matthew Hirn
Abstract:
The dream of machine learning in materials science is for a model to learn the underlying physics of an atomic system, allowing it to move beyond interpolation of the training set to the prediction of properties that were not present in the original training data. In addition to advances in machine learning architectures and training techniques, achieving this ambitious goal requires a method to c…
▽ More
The dream of machine learning in materials science is for a model to learn the underlying physics of an atomic system, allowing it to move beyond interpolation of the training set to the prediction of properties that were not present in the original training data. In addition to advances in machine learning architectures and training techniques, achieving this ambitious goal requires a method to convert a 3D atomic system into a feature representation that preserves rotational and translational symmetry, smoothness under small perturbations, and invariance under re-ordering. The atomic orbital wavelet scattering transform preserves these symmetries by construction, and has achieved great success as a featurization method for machine learning energy prediction. Both in small molecules and in the bulk amorphous $\text{Li}_α\text{Si}$ system, machine learning models using wavelet scattering coefficients as features have demonstrated a comparable accuracy to Density Functional Theory at a small fraction of the computational cost. In this work, we test the generalizability of our $\text{Li}_α\text{Si}$ energy predictor to properties that were not included in the training set, such as elastic constants and migration barriers. We demonstrate that statistical feature selection methods can reduce over-fitting and lead to remarkable accuracy in these extrapolation tasks.
△ Less
Submitted 16 July, 2020; v1 submitted 1 June, 2020;
originally announced June 2020.
-
Multiscale-Spectral GFEM and Optimal Oversampling
Authors:
Ivo Babuska,
Robert Lipton,
Paul Sinz,
Michael Stuebner
Abstract:
In this work we address the Multiscale Spectral Generalized Finite Element Method (MS-GFEM) developed in [I. Babuška and R. Lipton, Multiscale Modeling and Simulation 9 (2011), pp. 373--406]. We outline the numerical implementation of this method and present simulations that demonstrate contrast independent exponential convergence of MS-GFEM solutions. We introduce strategies to reduce the computa…
▽ More
In this work we address the Multiscale Spectral Generalized Finite Element Method (MS-GFEM) developed in [I. Babuška and R. Lipton, Multiscale Modeling and Simulation 9 (2011), pp. 373--406]. We outline the numerical implementation of this method and present simulations that demonstrate contrast independent exponential convergence of MS-GFEM solutions. We introduce strategies to reduce the computational cost of generating the optimal oversampled local approximating spaces used here. These strategies retain accuracy while reducing the computational work necessary to generate local bases. Motivated by oversampling we develop a nearly optimal local basis based on a partition of unity on the boundary and the associated A-harmonic extensions.
△ Less
Submitted 19 October, 2019;
originally announced October 2019.
-
Steerable Wavelet Scattering for 3D Atomic Systems with Application to Li-Si Energy Prediction
Authors:
Xavier Brumwell,
Paul Sinz,
Kwang ** Kim,
Yue Qi,
Matthew Hirn
Abstract:
A general machine learning architecture is introduced that uses wavelet scattering coefficients of an inputted three dimensional signal as features. Solid harmonic wavelet scattering transforms of three dimensional signals were previously introduced in a machine learning framework for the regression of properties of small organic molecules. Here this approach is extended for general steerable wave…
▽ More
A general machine learning architecture is introduced that uses wavelet scattering coefficients of an inputted three dimensional signal as features. Solid harmonic wavelet scattering transforms of three dimensional signals were previously introduced in a machine learning framework for the regression of properties of small organic molecules. Here this approach is extended for general steerable wavelets which are equivariant to translations and rotations, resulting in a sparse model of the target function. The scattering coefficients inherit from the wavelets invariance to translations and rotations. As an illustration of this approach a linear regression model is learned for the formation energy of amorphous lithium-silicon material states trained over a database generated using plane-wave Density Functional Theory methods. State-of-the-art results are produced as compared to other machine learning approaches over similarly generated databases.
△ Less
Submitted 28 January, 2019; v1 submitted 21 November, 2018;
originally announced December 2018.
-
Uncertain Loading and Quantifying Maximum Energy Concentration within Composite Structures
Authors:
Robert Lipton,
Paul Sinz,
Michael Stuebner
Abstract:
We introduce a systematic method for identifying the worst case load among all boundary loads of fixed energy. Here the worst case load is defined to be the one that delivers the largest fraction of input energy to a prescribed subdomain of interest. The worst case load is identified with the first eigenfunction of a suitably defined eigenvalue problem. The first eigenvalue for this problem is the…
▽ More
We introduce a systematic method for identifying the worst case load among all boundary loads of fixed energy. Here the worst case load is defined to be the one that delivers the largest fraction of input energy to a prescribed subdomain of interest. The worst case load is identified with the first eigenfunction of a suitably defined eigenvalue problem. The first eigenvalue for this problem is the maximum fraction of boundary energy that can be delivered to the subdomain. We compute worst case boundary loads and associated energy contained inside a prescribed subdomain through the numerical solution of the eigenvalue problem. We apply this computational method to bound the worst case load associated with an ensemble of random boundary loads given by a second order random process. Several examples are carried out on heterogeneous structures to illustrate the method.
△ Less
Submitted 2 August, 2016; v1 submitted 30 June, 2016;
originally announced July 2016.
