Role of inhibition and inhibitory synaptic plasticity

To understand how stable computation and memory emerge in plastic spiking neural networks, I am spending a lot of time thinking about stability of these systems. Because Hebbian plasticity needs to be kept in check by rapid negative feedback processes (see Online learning), I spent the better part of my PhD thinking about which rapid negative feedback mechanisms might exist in the brain.

At the beginning of my PhD with Wulfram I still thought that inhibitory plasticity alone could achieve this [1]. I am now quite confident that inhibitory plasticity alone is insufficient to stabilize Hebbian runaway dynamics unless it regulates synaptic plasticity directly for which we do not have a good model yet. However, Hebbian forms of inhibitory plasticity remain an essential mechanism to achieve and maintain excitatory and inhibitory balance in recurrent and feedfoward networks

[ref, Figure: Associative recall of a memory assembly in a spiking neural network in which the balanced state is established and maintained through inhibitory plasticity. Adapted from Vogels et al. 2011].

Functionally, it is appealing to think that this mainly serves a decorrelating purpose. However, different variations of inhibitory plasticity which could be subject to global modulatory factors might take important homeostatic roles in regulating the overall activity and excitability of populations of spiking neurons [3]. Because many details about different forms of inhibitory plasticity and their precise function remain elusive it remains an important direction which I am actively exploring in my research.

Supplementary Video

Here are some videos from our paper on inhibitory plasticity. This video is part of the Science press package. If you would like to show this material or parts of it, please provide a reference to our paper: Vogels, T.P., Sprekeler, H., Zenke, F., Clopath, C., Gerstner, W., 2011. Inhibitory Plasticity Balances Excitation and Inhibition in Sensory Pathways and Memory Networks. Science 334, 1569 –1573.

You can download the video here. Note, however, that this larger 250k cell network simulation is not part of the original publication. The simulation code is included in the recent release of the Auryn simulator. Reimplementations of the smaller 10k cell network are available in the example sections of Auryn, the Brian Simulator and on ModelDB (includes scripts to generate also the first figures of the paper).

Further Reading

  1. Zenke, F., Agnes, E.J., and Gerstner, W. (2015). Diverse synaptic plasticity mechanisms orchestrated to form and retrieve memories in spiking neural networks. Nat Commun 6.
  2. D’Amour, J.A., and Froemke, R.C. (2015). Inhibitory and Excitatory Spike-Timing-Dependent Plasticity in the Auditory Cortex. Neuron 86, 514–528. full-text
  3. Vogels, T.P., Froemke, R.C., Doyon, N., Gilson, M., Haas, J.S., Liu, R., Maffei, A., Miller, P., Wierenga, C., Woodin, M.A., Zenke, F., Sprekeler, H., 2013.
    Inhibitory Synaptic Plasticity – Spike timing dependence and putative network function. Front Neural Circuits 7. doi:10.3389/fncir.2013.00119
  4. Vogels, T.P., Sprekeler, H., Zenke, F., Clopath, C., Gerstner, W., 2011.
    Inhibitory Plasticity Balances Excitation and Inhibition in Sensory Pathways and Memory Networks. Science 334, 1569–1573. doi:10.1126/science.1211095