Neural mechanisms improving spike timing reliability under noisy and unreliable conditions
In this contribution we discuss a set of computational experiments which demonstrate that spiking neurons are able to learn and to precisely reproduce sequences of spikes even for highly noisy and unreliable input signals. As a result of our study we identified some cellular mechanisms which can potentially be exploited by the biological neurons to deal with noise and to produce accurate and reliable responses. Method: A single Leaky-Integrate-and-Fire neuron with multiple synaptic inputs (n=100) was repeatedly stimulated with a set of Poissonian stimuli. The neuron was trained to respond with some predefined sequences of spikes to the particular sets of input signals. The training was performed according to ReSuMe - a supervised learning algorithm operating on precise spikes timing [1]. In the particular experiments different sources of noise were simulated by adding a background or synaptic noise to the neuron or by introducing a jitter to the spike timing of the input and target signals. In each experiment the sequences of spikes generated by the neuron were recorded. The correlation of the recorded patterns with the target sequences as well as the variability of the generated spikes over the particular trials were calculated. Results and conclusions: Our results demonstrate accurate and reliable spike timing in response to the stimuli used during training. At the same time it is observed that the neuron spikes highly unreliably for other stimuli (not used during training). We note, however, that the necessary condition in our experiments for the neuron's reliability was that the training was performed already in a presence of noise. In a case of 'noise-free training', the reliability of neuron in the noisy testing phase was reduced dramatically. Our findings are consistent with the experimental results, which indicate that the same neuron may have very accurate spike timing in response to one stimulus and unreliable spike timing for another [2]. This observation suggests also that the results of at least some in vivo and in vitro experiments revealing the unreliability of spike timing of the biological neurons should be reinvestigated, since they could erroneously be affected by use of inappropriate stimuli. As a result of our study we identified some mechanisms by which the spiking neurons can deal with noise and unreliability. These are: (1) significant increase of excitation shortly before the target firing times, and the reduction of excitation at all other times; (2) reduced impact of the individual EPSPs on the spike generation in favour of the groups of EPSPs. The identified mechanisms are not necessary attributed to the particular properties of the learning method used in our experiment and the same effects are likely to arise while using the learning rules possibly utilized in the central nervous system. Hence, they are proposed as potential candidates for the mechanisms employed by the biological neurons to deal with noise.
References
1. F.Ponulak, Supervised Learning in Spiking Neural Networks with ReSuMe Method, PhD Thesis, Poznan University of Technology, Poland, 2006
2. E.Schneidman, Noise and Information in Neural Codes. PhD Thesis, The Hebrew University, Israel, 2001.
Conference:
Bernstein Symposium 2008, Munich, Germany, 8 Oct - 10 Oct, 2008.
Presentation Type:
Poster Presentation
Topic:
All Abstracts
Citation:
Ponulak
F
(2008). Neural mechanisms improving spike timing reliability under noisy and unreliable conditions.
Front. Comput. Neurosci.
Conference Abstract:
Bernstein Symposium 2008.
doi: 10.3389/conf.neuro.10.2008.01.105
Copyright:
The abstracts in this collection have not been subject to any Frontiers peer review or checks, and are not endorsed by Frontiers.
They are made available through the Frontiers publishing platform as a service to conference organizers and presenters.
The copyright in the individual abstracts is owned by the author of each abstract or his/her employer unless otherwise stated.
Each abstract, as well as the collection of abstracts, are published under a Creative Commons CC-BY 4.0 (attribution) licence (https://creativecommons.org/licenses/by/4.0/) and may thus be reproduced, translated, adapted and be the subject of derivative works provided the authors and Frontiers are attributed.
For Frontiers’ terms and conditions please see https://www.frontiersin.org/legal/terms-and-conditions.
Received:
17 Nov 2008;
Published Online:
17 Nov 2008.
*
Correspondence:
Filip Ponulak, BCCN Freiburg, Freiburg, Germany, ponulak@bccn.uni-freiburg.de