Visible to the public Response Time Analysis for Explainability of Visual Processing in CNNs

TitleResponse Time Analysis for Explainability of Visual Processing in CNNs
Publication TypeConference Paper
Year of Publication2020
AuthorsTaylor, E., Shekhar, S., Taylor, G. W.
Conference Name2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW)
KeywordsAnalytical models, CNNs, cognitive psychology, Computational modeling, conditional computation model, convolutional neural nets, dynamic inference models, early-exit architecture, explainable artificial intelligence methods, Grammar, hierarchical representations, intrascene object-object relationships, learning (artificial intelligence), model architecture, MSDNet, Object recognition, pubcrawl, Resiliency, response times, Scalability, Semantics, Syntactics, variable RTs, visual learning tasks, visual processing, visualization, xai
AbstractExplainable artificial intelligence (XAI) methods rely on access to model architecture and parameters that is not always feasible for most users, practitioners, and regulators. Inspired by cognitive psychology, we present a case for response times (RTs) as a technique for XAI. RTs are observable without access to the model. Moreover, dynamic inference models performing conditional computation generate variable RTs for visual learning tasks depending on hierarchical representations. We show that MSDNet, a conditional computation model with early-exit architecture, exhibits slower RT for images with more complex features in the ObjectNet test set, as well as the human phenomenon of scene grammar, where object recognition depends on intrascene object-object relationships. These results cast light on MSDNet's feature space without opening the black box and illustrate the promise of RT methods for XAI.
Citation Keytaylor_response_2020