Please use this identifier to cite or link to this item:
https://scholarbank.nus.edu.sg/handle/10635/153716
DC Field | Value | |
---|---|---|
dc.title | LAYERED EXPLANATIONS - INTERPRETING NEURAL NETWORKS WITH NUMERICAL INFLUENCE MEASURES | |
dc.contributor.author | HO XUAN VINH | |
dc.date.accessioned | 2019-05-06T18:01:35Z | |
dc.date.available | 2019-05-06T18:01:35Z | |
dc.date.issued | 2019-01-10 | |
dc.identifier.citation | HO XUAN VINH (2019-01-10). LAYERED EXPLANATIONS - INTERPRETING NEURAL NETWORKS WITH NUMERICAL INFLUENCE MEASURES. ScholarBank@NUS Repository. | |
dc.identifier.uri | https://scholarbank.nus.edu.sg/handle/10635/153716 | |
dc.description.abstract | Deep learning is currently receiving considerable attention from the machine learning community due to its predictive power. However, its lack of interpretability raises numerous concerns. Since neural networks are deployed in high-stakes domains, stakeholders expect to receive acceptable human interpretable explanations. We explain the decisions of neural networks using layered explanations: we use influence measures in order to compute a numerical value for each layer. Using layerwise influence measures, we identify the layers that contain the most explanatory power, and use those to generate explanations. We test our methodology on datasets, and discuss the merits and issues with our approach. | |
dc.language.iso | en | |
dc.subject | algorithmic transparency, interpretable explanation, model-specific explanation, numerical influence measures, neural networks, deep learning | |
dc.type | Thesis | |
dc.contributor.department | COMPUTER SCIENCE | |
dc.contributor.supervisor | YAIR ZICK | |
dc.description.degree | Master's | |
dc.description.degreeconferred | MASTER OF SCIENCE | |
Appears in Collections: | Master's Theses (Open) |
Show simple item record
Files in This Item:
File | Description | Size | Format | Access Settings | Version | |
---|---|---|---|---|---|---|
HoXV.pdf | 4.71 MB | Adobe PDF | OPEN | None | View/Download |
Google ScholarTM
Check
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.