Device for machine learning molecular dynamics using error function
According to an embodiment, an information processing device includes one or more hardware processors configured to: set an error function including one or more terms based on a plurality of weights according to features of a plurality of elements, the error function being a function used during learning of a machine learning model into which positions of a plurality of atoms included in an analysis target, and information indicating which of the plurality of elements the plurality of atoms are, are input, and that outputs a physical quantity of the analysis target; and learn the machine learning model using the error function.
1 . An information processing device comprising
one or more hardware processors configured to:
set an error function including one or more terms based on a plurality of weights according to features of a plurality of elements, the error function being a function used during learning of a machine learning model into which positions of a plurality of atoms included in an analysis target, and information indicating which of the plurality of elements the plurality of atoms are, are input, and that outputs a physical quantity of the analysis target; and
learn the machine learning model using the error function,
wherein the one or more hardware processors are configured to set the error function including a plurality of terms obtained by multiplying, by the plurality of weights, errors of physical quantities for the plurality of elements.
2 . The device according to claim 1 ,
wherein the one or more hardware processors are configured to:
set the plurality of weights according to the features of the plurality of elements, and
set the error function including the one or more terms based on the plurality of set weights.
3 . The device according to claim 2 ,
wherein the one or more hardware processors are configured to:
aggregate numbers of atoms that are numbers of atoms of the plurality of elements included in a plurality of pieces of learning data used for learning of the machine learning model, for the plurality of elements, based on the plurality of pieces of learning data, the plurality of pieces of learning data each including positions of the plurality of atoms, and
set the plurality of weights based on the numbers of atoms or ratios of the numbers of atoms to a total number of atoms of all elements included in the plurality of pieces of learning data.
4 . The device according to claim 1 ,
wherein the one or more hardware processors are configured to:
receive the plurality of weights designated for the plurality of elements, and
set the error function including the one or more terms based on the plurality of received weights.
5 . The device according to claim 1 ,
wherein the features are masses of the plurality of elements, atomic numbers of the plurality of elements, information correlated with the masses or the atomic numbers, or numbers of atoms of the plurality of elements included in the analysis target.
6 . The device according to claim 5 ,
wherein the plurality of weights are reciprocals of the plurality of features.
7 . The device according to claim 1 ,
wherein the one or more hardware processors are configured to set the error function including a plurality of terms obtained by multiplying second errors by the weights, the second errors being sums of first errors that are errors of the physical quantities of the plurality of atoms, for the plurality of elements.
8 . The device according to claim 1 ,
wherein the one or more hardware processors are configured to select one or more elements from among the plurality of elements with a probability according to magnitude of the plurality of weights, and set the error function including one or more terms obtained by multiplying weights for the selected one or more elements by errors of the physical quantities for the selected one or more elements.
9 . The device according to claim 8 ,
wherein the one or more hardware processors are configured to select the one or more elements with the probabilities for each of a plurality of pieces of learning data used for learning of the machine learning mode, and each including positions of the plurality of atoms.
10 . The device according to claim 8 ,
wherein the one or more hardware processors are configured to:
repeatedly execute learning of the machine learning model a plurality of times using a plurality of pieces of learning data each including positions of the plurality of atoms; and
select one or more of the plurality of elements with the probabilities for each of the plurality of times of the learning.
11 . The device according to claim 1 ,
wherein the one or more hardware processors are configured to:
correct at least part of the plurality of weights to a designated value, and
set the error function including one or more terms based on the plurality of weights after correction.
12 . The device according to claim 11 ,
wherein the one or more hardware processors are configure to:
output an output value of the error function; and
correct at least part of the plurality of weights to a value designated according to the output output value.
13 . The device according to claim 1 ,
wherein the one or more hardware processors are configured to correct a parameter of the machine learning model such that an output value of the error function is made small.
14 . The device according to claim 1 ,
wherein the one or more hardware processors are configured to implement:
a function setting unit configured to set the error function; and
a learning unit configured to learn the machine learning model.
15 . An information processing method executed by an information processing device, comprising:
setting an error function including one or more terms based on a plurality of weights according to features of a plurality of elements, the error function being a function used during learning of a machine learning model into which positions of a plurality of atoms included in an analysis target, and information indicating which of the plurality of elements the plurality of atoms are, are input, and that outputs a physical quantity of the analysis target; and
learning the machine learning model using the error function; and
wherein the setting the error function including a plurality of terms obtained by multiplying, by the plurality of weights, errors of physical quantities for the plurality of elements.
16 . A computer program product comprising a non-transitory computer-readable medium including programmed instructions, the instructions causing a computer to execute:
setting an error function including one or more terms based on a plurality of weights according to features of a plurality of elements, the error function being a function used during learning of a machine learning model into which positions of a plurality of atoms included in an analysis target, and information indicating which of the plurality of elements the plurality of atoms are, are input, and that outputs a physical quantity of the analysis target; and
learning the machine learning model using the error function; and
wherein the setting the error function including a plurality of terms obtained by multiplying, by the plurality of weights, errors of physical quantities for the plurality of elements.