Predicting non-linear dynamics by stable local learning in a recurrent spiking neural network
- Submitting institution
-
The University of Sheffield
- Unit of assessment
- 11 - Computer Science and Informatics
- Output identifier
- 5648
- Type
- D - Journal article
- DOI
-
10.7554/elife.28295
- Title of journal
- eLife
- Article number
- e28295
- First page
- -
- Volume
- 6
- Issue
- -
- ISSN
- 2050-084X
- Open access status
- Compliant
- Month of publication
- November
- Year of publication
- 2017
- 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)
-
C - Machine Learning
- Citation count
- 23
- Proposed double-weighted
- No
- Reserve for an output with double weighting
- No
- Additional information
- This is the first paper to use adaptive control theory to derive and prove stability and convergence of a biologically-plausible learning scheme that enables recurrent spiking neural networks to predict non-linear body dynamics. This scheme is implementable on low-power neuromorphic hardware, for which discussion is ongoing with the world-leading group of Professor Indiveri (INI Zurich). The article was recommended at the "Faculty of 1000 Prime" (now Faculty Opinions https://facultyopinions.com/prime/732181243). Our follow-up work on bio-plausible learning for motor control was highlighted as a “long talk” (https://icml.cc/Conferences/2018/Schedule?showEvent=2671) at the premier International Conference on Machine Learning (ICML) 2018 (http://proceedings.mlr.press/v80/gilra18a.html).
- Author contribution statement
- -
- Non-English
- No
- English abstract
- -