What do deep neural networks tell us about biological vision?
• 大类 : 医学 - 4区
• 小类 : 神经科学 - 4区
• 小类 : 眼科学 - 3区
• 小类 : 心理学 - 4区
Recent years have seen a huge increase in the application of deep learning techniques and ‘biologically inspired’ deep neural networks (DNNs) to a broad range of issues in biological vision. Indeed, DNNs have been described by some as a new framework for vision research, allowing an opportunity to ‘reverse engineer’ the biological system. These claims are, in part, based on work showing human-level performance by DNNs in tasks such as image classification and are supported by advances in the development of methods for comparing representational structures computed by DNNs with biological vision systems. But the suitability of such networks as a theoretical framework for understanding biological vision is unclear. There remain many important questions: How should theoretically relevant and irrelevant properties of DNN architectures and processing parameters be distinguished? How can network performance be rigorously compared with corresponding biological data? What is the range of relevant performance data for evaluating network outputs? And to what extent can network activity be used to formulate empirically testable models of biological vision? This special issue invites novel contributions on these and related topics. We welcome original articles that consider the application of DNNs to understanding any aspect of biological vision.