S
Stephen Casper
Researcher at Harvard University
Publications - 19
Citations - 73
Stephen Casper is an academic researcher from Harvard University. The author has contributed to research in topics: Computer science & Dialog box. The author has an hindex of 4, co-authored 11 publications receiving 23 citations. Previous affiliations of Stephen Casper include Boston Children's Hospital & Massachusetts Institute of Technology.
Papers
More filters
Journal ArticleDOI
Toward Transparent AI: A Survey on Interpreting the Inner Structures of Deep Neural Networks
TL;DR: In this survey, literature on techniques for interpreting the inner components of DNNs, which are called inner interpretability methods are reviewed, with a focus on how these techniques relate to the goal of designing safer, more trustworthy AI systems.
Proceedings ArticleDOI
Probing Neural Dialog Models for Conversational Understanding
TL;DR: It is suggested that standard open-domain dialog systems struggle with answering questions, inferring contradiction, and determining the topic of conversation, among other tasks, and the dyadic, turn-taking nature of dialog is not fully leveraged by these models.
Posted Content
Frivolous Units: Wider Networks Are Not Really That Wide
Stephen Casper,Xavier Boix,Vanessa D'Amario,Ling Guo,Martin Schrimpf,Kasper Vinken,Gabriel Kreiman +6 more
TL;DR: This work identifies two distinct types of "frivolous" units that proliferate when the network's width is increased: prunable units which can be dropped out of the network without significant change to the output and redundant units whose activities can be expressed as a linear combination of others.
Posted Content
Clusterability in Neural Networks.
TL;DR: In this article, the authors look for structure in the form of clusterability: how well a network can be divided into groups of neurons with strong internal connectivity but weak external connectivity and find that a trained neural network is typically more clusterable than randomly initialized networks, and often clusterable relative to random networks with the same distribution of weights.
Journal ArticleDOI
Benchmarking Interpretability Tools for Deep Neural Networks
TL;DR: In this paper , the authors propose trojan rediscovery as a benchmarking task to evaluate how useful interpretability tools are for generating engineering-relevant insights, and apply their benchmarks to evaluate 16 feature attribution/saliency methods and 9 feature synthesis methods.