Imperial College London
Alternate

Dr Eric E Keaveny

Faculty of Natural Sciences, Department of Mathematics

Reader in Applied Mathematics
 
 
 
//

Contact

 

+44 (0)20 7594 2780e.keaveny

 
 
//

Location

 

741Huxley BuildingSouth Kensington Campus

//

Summary

 

Publications

Citation

BibTex format

@inproceedings{Li:2018,
author = {Li, K and Javer, A and Keaveny, E and Brown, AE},
title = {Recurrent Neural Networks with Interpretable Cells Predict and Classify Worm Behaviour},
url = {http://hdl.handle.net/10044/1/56078},
year = {2018}
}
Download

RIS format (EndNote, RefMan)

TY  - CPAPER
AB  - An important goal in behaviour analytics is to connect disease state or genomevariation with observable differences in behaviour. Despite advances in sensortechnology and imaging, informative behaviour quantification remains challenging.The nematode worm C. elegans provides a unique opportunity to test analysisapproaches because of its small size, compact nervous system, and the availabilityof large databases of videos of freely behaving animals with known genetic differences.Despite its relative simplicity, there are still no reports of generative modelsthat can capture essential differences between even well-described mutant strains.Here we show that a multilayer recurrent neural network (RNN) can produce diversebehaviours that are difficult to distinguish from real worms’ behaviour andthat some of the artificial neurons in the RNN are interpretable and correlate withobservable features such as body curvature, speed, and reversals. Although theRNN is not trained to perform classification, we find that artificial neuron responsesprovide features that perform well in worm strain classification.
AU  - Li,K
AU  - Javer,A
AU  - Keaveny,E
AU  - Brown,AE
PY  - 2018///
TI  - Recurrent Neural Networks with Interpretable Cells Predict and Classify Worm Behaviour
UR  - http://hdl.handle.net/10044/1/56078
ER  -
Download