期刊
ISCIENCE
卷 24, 期 9, 页码 -出版社
CELL PRESS
DOI: 10.1016/j.isci.2021.103013
关键词
-
资金
- New Energy and Industrial Technology Development Organization (NEDO) [JPNP20006]
- JSPS KAKENHI [JP15H05920, JP15H05710, JP20H05705, 20H05954]
- Grants-in-Aid for Scientific Research [20H05954] Funding Source: KAKEN
Recent studies show that high-performance DNNs are not necessarily brain-like, and a single-path sequential feedforward architecture is crucial for brain-like hierarchy. By quantifying the correspondence between DNN unit activations and human brain activity through neural decoding and encoding analyses, new ways to design DNNs based on their representational homology to the brain may be provided.
Achievement of human-level image recognition by deep neural networks (DNNs) has spurred interest in whether and how DNNs are brain-like. Both DNNs and the visual cortex perform hierarchical processing, and correspondence has been shown between hierarchical visual areas and DNN layers in representing visual features. Here, we propose the brain hierarchy (BH) score as a metric to quantify the degree of hierarchical correspondence based on neural decoding and encoding analyses where DNN unit activations and human brain activity are predicted from each other. We find that BH scores for 29 pre-trained DNNs with various architectures are negatively correlated with image recognition performance, thus indicating that recently developed high-performance DNNs are not necessarily brain-like. Experimental manipulations of DNN models suggest that single-path sequential feedforward architecture with broad spatial integration is critical to brain-like hierarchy. Our method may provide new ways to design DNNs in light of their representational homology to the brain.
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