Timothée Masquelier

8 février 2013

Computational Neuroscience Group, Univ. Pompeu Fabra, Barcelona, Spain

Spike-based computing and learning in brains, machines, and visual systems in particular

invité par Laurent Perrinet

Abstract : Using simulations, we have first shown that, thanks to the physiological learning mechanism referred to as Spike Timing-Dependent Plasticity (STDP), neurons can detect and learn repeating spike patterns, in an unsupervised manner, even when those patterns are embedded in noise1,2. Importantly, the spike patterns do not need to repeat exactly : it also works when only a firing probability pattern repeats, providing this profile has narrow (10-20ms) temporal peaks3. Brain oscillations may help in getting the required temporal precision4, in particular when dealing with slowly changing stimuli. All together, these studies show that some envisaged problems associated to spike timing codes, in particular noise-resistance, the need for a reference time, or the decoding issue, might not be as severe as once thought.

These generic STDP-based learning mechanisms might be at work in particular in the visual system, where repeating spike patterns correspond to visual primitives. First, in an attempt to model ultra-rapid visual categorization with a feedforward network, we have found that a combination of a temporal coding scheme where in the input layers the most strongly activated neurons fire first, with STDP leads to a situation where neurons in subsequent layers will gradually become selective to prototypical visual patterns that are both salient and consistently present in the images5. At the same time, their responses become more and more rapid. We firmly believe that such mechanisms are not only a key to understanding the remarkable efficiency of the primate visual system, but also appealing for artificial vision systems.

More recently, we have explored continuous vision with a detailed model of the cat early visual system6. We used natural videos as inputs and converted them to spikes using the Virtual Retina simulator7. We then showed that downstream simple cells in V1, if equipped with STDP, tended to connect principally to ON- and OFF-centre cells with receptive fields aligned in the visual space – because those have correlated spike times – and thereby become orientation selective, in accordance with Hubel and Wiesel classic model8. It is worth mentioning that there was no absolute time reference such as a stimulus onset, yet information was encoded and decoded in the relative spike times.

Together with Linares’ group, we are implementing these model in hardware9, leading to simulations running orders of magnitude faster than real time. We speculate that this line of research will yield revolutionary results in the next decade.

References 1. Masquelier, T., Guyonneau, R. & Thorpe, S. J. Spike timing dependent plasticity finds the start of repeating patterns in continuous spike trains. PLoS ONE 3, e1377 (2008).

2. Masquelier, T., Guyonneau, R. & Thorpe, S. J. Competitive STDP-Based Spike Pattern Learning. Neural Comput 21, 1259–1276 (2009).

3. Gilson, M., Masquelier, T. & Hugues, E. STDP allows fast rate-modulated coding with Poisson-like spike trains. PLoS Comput Biol 7, e1002231 (2011).

4. Masquelier, T., Hugues, E., Deco, G. & Thorpe, S. J. Oscillations, phase-of-firing coding, and spike timing- dependent plasticity : an efficient learning scheme. J Neurosci 29, 13484–93 (2009).

5. Masquelier, T. & Thorpe, S. J. Unsupervised learning of visual features through spike timing dependent plasticity. PLoS Comput Biol 3, e31 (2007).

6. Masquelier, T. Relative spike time coding and STDP-based orientation selectivity in the early visual system in natural continuous and saccadic vision : a computational model. J Comput Neurosci 32, 425–41 (2012).

7. Wohrer, A. & Kornprobst, P. Virtual Retina : a biological retina model and simulator, with contrast gain control. J Comput Neurosci 26, 219–249 (2009).

8. Hubel, D. H. & Wiesel, T. N. Receptive fields, binocular interaction and functional architecture in the cat’s visual cortex. J Physiol 160, 106–154 (1962).

9. Zamarreño-Ramos, C. et al. On Spike-Timing-Dependent-Plasticity, Memristive Devices, and Building a Self-Learning Visual Cortex. Frontiers Neurosci 5, 22 (2011).

CNRS logo université Aix Marseille logo | plan du site | mentions légales | contact | admin | intranet | intcloud |