Simulation device, simulation method, and memory medium

A simulation method, the method comprising:

• when an initial state of a state vector and a parameter in a simulation and a plurality of pieces of observation data are input,
• simulating a time evolution of the state vector using a system model based on the initial state and the parameter;
• selecting, from the plurality of pieces of observation data, a plurality of pieces of observation data to be used based on information related to the state vector in the system model;
• transforming, by use of a plurality of observation models each of which is associated with one of the selected plurality of pieces of observation data, the state vector output from the system model based on a relationship between the piece of observation data and the state vector;
• creating posterior distributions of the state vector based on state vectors output from the plurality of observation models and the selected pieces of observation data;
• performing unification of a first posterior distribution based on all the selected pieces of observation data and a second posterior distribution based on a set of observation data lacking one or more pieces of observation data;
• determining which one of the second posterior distribution and a posterior distribution after the unification is to be used;
• inputting a state vector including a determined posterior distribution and the first posterior distribution to the system model;
• and
• outputting a time series of a state vector including a determined posterior distribution and the first posterior distribution
• wherein the state vector represents a subject, and
• wherein the observation data is obtained by different sensors measuring the subject.

A non-transitory computer-readable storage medium storing a computer program, the program making a computer device execute:

• input processing of obtaining an initial state of a state vector and a parameter in a simulation and a plurality of pieces of observation data as input;
• system model calculation processing of, based on the initial state and the parameter, simulating a time evolution of the state vector using a system model;
• data selection processing of, based on information relating to the state vector in the system model, selecting, from the plurality of pieces of observation data, a plurality of pieces of observation data to be used;
• observation model calculation processing of, by use of a plurality of observation models each of which is associated with one of the selected plurality of pieces of observation data, transforming and outputting each state vector output from the system model based on a relationship between the piece of observation data and the state vector;
• posterior distribution creating processing of, based on state vectors output from the plurality of observation models and pieces of observation data selected in the data selection processing, creating posterior distributions of the state vector, outputting a posterior distribution based on all pieces of observation data selected in the data selection processing as a first posterior distribution, and outputting a posterior distribution based on a set of observation data lacking one or more pieces of observation data as a second posterior distribution;
• posterior distribution unifying processing of performing unification of the first posterior distribution and the second posterior distribution;
• determining processing of determining which one of the second posterior distribution and a posterior distribution after the unification is to be used;
• and
• output processing of inputting a state vector including a posterior distribution determined in the determining processing and the first posterior distribution to the system model and outputting a time series of the state vector,
• wherein the state vector represents a subject, and
• wherein the observation data is obtained by different sensors measuring the subject.