基于优势决策的神经机制解码方法研究

Decoding method of neural mechanism based on advantageous decision making

研究了优势决策中的神经机制解码方法。神经机制解码方法分为基于统计分析的单变量分析和基于机器学习的多变量分析。通过比较基于单变量的广义线性模型和基于多变量的多体素模式分析,探索优势决策中的解码方法。单变量分析中,采用单特征的逻辑斯蒂回归建立模型,多变量分析中,对比了以K近邻、朴素贝叶斯、线性支持向量机、非线性支持向量机为分类器的分类模型。结果表明利用线性支持向量机建立的模型,可以得到优于其他方法的解码精度,且多变量分类模型解码能力均在统计意义上优于单变量模型。

The paper studies the decoding method of neural activation in advantageous decision making.The neural activation decoding method is divided into univariate analysis based on statistical analysis and multivariate analysis based on machine learning.The paper explores the decoding method in advantageous decision making by comparing the generalized linear model (GLM) based on single variables and the multivariable multi-voxel pattern analysis(MVPA).In the univariate analysis,the paper uses a single feature logistic regression to obtain a model.In multivariate analysis,the K-nearest neighbor,naive Bayesian,linear support vector machine,and nonlinear support vector machine are used as the classifier classification model.The result shows that the model trained by linear support vector machine can perform better than other methods,and the decoding ability of multivariate model is significant better than the univariate model.