期刊
NEUROIMAGE
卷 137, 期 -, 页码 188-200出版社
ACADEMIC PRESS INC ELSEVIER SCIENCE
DOI: 10.1016/j.neuroimage.2015.12.012
关键词
fMRI; Multi-voxel pattern analysis; Representational similarity analysis; Crossvalidation; Linear discriminant; Noise normalization; Classification; Decoding; Machine learning
资金
- Gates Cambridge Scholarship
- European Research Council Starting Grant [ERC-2010-StG 261352]
- Wellcome Trust [094874/Z/10/Z, 091593/Z/10/Z, WT091540MA]
- Wellcome Trust [094874/Z/10/Z] Funding Source: Wellcome Trust
- Medical Research Council [MC_U105597120] Funding Source: researchfish
- MRC [MC_U105597120] Funding Source: UKRI
Representational similarity analysis of activation patterns has become an increasingly important tool for studying brain representations. The dissimilarity between two patterns is commonly quantified by the correlation distance or the accuracy of a linear classifier. However, there are many different ways to measure pattern dissimilarity and little is known about their relative reliability. Here, we compare the reliability of three classes of dissimilarity measure: classification accuracy, Euclidean/Mahalanobis distance, and Pearson correlation distance. Using simulations and four real functional magnetic resonance imaging (fMRI) datasets, we demonstrate that continuous dissimilarity measures are substantially more reliable than the classification accuracy. The difference in reliability can be explained by two characteristics of classifiers: discretization and susceptibility of the discriminant function to shifts of the pattern ensemble between imaging runs. Reliability can be further improved through multivariate noise normalization for all measures. Finally, unlike conventional distance measures, crossvalidated distances provide unbiased estimates of pattern dissimilarity on a ratio scale, thus providing an interpretable zero point. Overall, our results indicate that the crossvalidated Mahalanobis distance is preferable to both the classification accuracy and the correlation distance for characterizing representational geometries. (C) 2015 Published by Elsevier Inc.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据