Method and apparatus for training causal model

Abstract

Embodiments of the disclosure relate to a method and apparatus for training a causal model, and a computer readable storage medium. For example, the method for training the causal model includes steps: establishing the causal model based on a plurality of observation variables and at least one hidden variable, wherein the causal model comprises a first parameter and a second parameter which are tobe determined, the first parameter indicates first relations between the plurality of observation variables, and the second parameter indicates second relations between the at least one hidden variable and the plurality of observation variables; determining the second parameter and a third parameter associated with the first parameter by employing a probability principal component analysis; determining noise of the causal model based on the second parameter and the third parameter; and determining the first parameter based on the noise. The embodiment of the disclosure also provides an apparatus capable of realizing the above method and a computer readable storage medium.

Description

technical field [0001] Embodiments of the present disclosure relate to the field of machine learning, and more particularly, to methods, devices, and computer-readable storage media for training causal models. Background technique [0002] With the rapid development of information technology, the scale of data is growing rapidly. Against such a background and trend, machine learning has received more and more attention. Among them, causality discovery (such as linear causality discovery, linear latent variable causality discovery, etc.) has a wide range of applications in real life, such as supply chain, medical health and retail and other fields. However, due to the existence of hidden variables and the unknown effects of hidden variables on observed variables, solving linear causality involving hidden variables is an important and difficult challenge in causal discovery. [0003] Some traditional schemes can use the method of complete independent component analysis to fi...

AI Technical Summary

Problems solved by technology

However, since such methods employ scaled conjugate gradient methods for parameter estimation, they are not efficient for linear causality and cannot determine causal structure against Gaussian noise

Images