Towards explainable deep neural networks (xDNN)
- Submitting institution
-
The University of Lancaster
- Unit of assessment
- 11 - Computer Science and Informatics
- Output identifier
- 302399780
- Type
- D - Journal article
- DOI
-
10.1016/j.neunet.2020.07.010
- Title of journal
- Neural Networks
- Article number
- -
- First page
- 185
- Volume
- 130
- Issue
- -
- ISSN
- 0893-6080
- Open access status
- Compliant
- Month of publication
- July
- Year of publication
- 2020
- URL
-
-
- Supplementary information
-
-
- Request cross-referral to
- -
- Output has been delayed by COVID-19
- No
- COVID-19 affected output statement
- -
- Forensic science
- No
- Criminology
- No
- Interdisciplinary
- No
- Number of additional authors
-
1
- Research group(s)
-
B - Data Science
- Citation count
- 2
- Proposed double-weighted
- No
- Reserve for an output with double weighting
- No
- Additional information
- This paper is the first explainable-by-design deep neural network classifier. The proposed approach is prototype-based and offers human-friendly visualisation and linguistic (rule-based) explanations unlike other existing methods that offer surrogate models or feature-based analysis. The proposed approach is unique in bringing together reasoning and machine learning. It offers better results and faster and easier training and was applied to various problems, e.g. to Covid19 diagnostics, driverless cars and remote sensing. This work led to an European Space Agency funded project (180K Euro).
- Author contribution statement
- -
- Non-English
- No
- English abstract
- -
