초록

용어보기논문에서 용어와 풀이말을 자동 추출한 결과로, 시범 서비스 중입니다. AI-Helper

Abstract ▼ AI-Helper

Here we propose an evolutionary method for learning hypernetworks based on sequential Bayesian sampling for high-dimensional data. The hypernetworks is learned to maximize the conditional posterior distribution with the proposed method by defining the model fitness to the posterior. The prior distri...

Here we propose an evolutionary method for learning hypernetworks based on sequential Bayesian sampling for high-dimensional data. The hypernetworks is learned to maximize the conditional posterior distribution with the proposed method by defining the model fitness to the posterior. The prior distribution is defined to the function of both prior knowledge of the problem and the model complexity for efficient learning. The likelihood is estimated by measuring the discriminative capability on the given data of the model. Therefore, the hypernetworks can efficiently learn very high-dimensional data and be improved with respect to the classification performance and the learning time by our method. Moreover, significant parameters in learning hypernetworks such as hyperedge degrees and the composition of a hyperedge are adaptively determined by the fitness. For evaluation, we apply the proposed method to classifying the two types of prostate cancer from expression dataset consisting of approximately 25,000 genes. Experimental results show that the hypernetwork applying the proposed method outperforms conventional hypernetworks as well as other classification models. Also, we present how the information used as the prior distribution influences on the model learning with diverse experiments.

Keyword