Hebbian plasticity requires compensatory processes on multiple timescales

We review a body of theoretical and experimental research on Hebbian and homeostatic plasticity, starting from a puzzling observation: while homeostasis of synapses found in experiments is a slow compensatory process, most mathematical models of synaptic plasticity use rapid compensatory processes (RCPs). Even worse, with the slow homeostatic plasticity reported in experiments, simulations of existing plasticity models cannot maintain network stability unless further control mechanisms are implemented. To solve this paradox, we suggest that in addition to slow forms of homeostatic plasticity there are RCPs which stabilize synaptic plasticity on short timescales. These rapid processes may include heterosynaptic depression triggered by episodes of high postsynaptic firing rate. While slower forms of homeostatic plasticity are not sufficient to stabilize Hebbian plasticity, they are important for fine-tuning neural circuits. Taken together we suggest that learning and memory rely on an intricate interplay of diverse plasticity mechanisms on different timescales which jointly ensure stability and plasticity of neural circuits. This article is part of the themed issue 'Integrating Hebbian and homeostatic plasticity'.

Memory formation and recall in recurrent spiking neural networks

Friedemann Zenke

Our brain has the capacity to analyze a visual scene in a split second, to learn how to play an instrument, and to remember events, faces and concepts. Neurons underlie all of these diverse functions. Neurons, cells within the brain that generate and transmit electrical activity, communicate with each other through chemical synapses. These synaptic connections dynamically change with experience, a process referred to as synaptic plasticity, which is thought to be at the core of the brain's ability to learn and process the world in sophisticated ways. Our understanding of the rules of synaptic plasticity remains quite limited. To enable efficient computations among neurons or to serve as a trace of memory, synapses must create stable connectivity patterns between neurons. However there remains an insufficient theoretical explanation as to how stable connectivity patterns can form in the presence of synaptic plasticity. Since the dynamics of recurrently connected neurons depend upon their connections, which themselves change in response to the network dynamics, synaptic plasticity and network dynamics have to be treated as a compound system. Due to the nonlinear nature of the system this can be analytically challenging. Utilizing network simulations that model the interplay between the network connectivity and synaptic plasticity can provide valuable insights. However, many existing network models that implement biologically relevant forms of plasticity become unstable. This suggests that current models do not accurately describe the biological networks, which have no difficulty functioning without succumbing to exploding network activity. The instability in these network simulations could originate from the fact that theoretical studies have, almost exclusively, focused on Hebbian plasticity at excitatory synapses. Hebbian plasticity causes connected neurons that are active together to increase the connection strength between them. Biological networks, however, display a large variety of different forms of synaptic plasticity and homeostatic mechanisms, beyond Hebbian plasticity. Furthermore, inhibitory cells can undergo synaptic plasticity as well. These diverse forms of plasticity are active at the same time, and our understanding of the computational role of most of these synaptic dynamics remains elusive. This raises the important question as to whether forms of plasticity that have not been previously considered could -in combination with Hebbian plasticity- lead to stable network dynamics. Here we illustrate that by combining multiple forms of plasticity with distinct roles, a recurrently connected spiking network model self-organizes to distinguish and extract multiple overlapping external stimuli. Moreover we show that the acquired network structures remain stable over hours while plasticity is active. This long-term stability allows the network to function as an associative memory and to correctly classify distorted or partially cued stimuli. During intervals in which no stimulus is shown the network dynamically remembers the last stimulus as selective delay activity. Taken together this work suggest that multiple forms of plasticity and homeostasis on different timescales have to work together to create stable connectivity patterns in neuronal networks which enable them to perform relevant computation.

EPFL2014

Synaptic Plasticity in Neural Networks Needs Homeostasis with a Fast Rate Detector

Wulfram Gerstner, Guillaume Hennequin, Friedemann Zenke

Hebbian changes of excitatory synapses are driven by and further enhance correlations between pre- and postsynaptic activities. Hence, Hebbian plasticity forms a positive feedback loop that can lead to instability in simulated neural networks. To keep activity at healthy, low levels, plasticity must therefore incorporate homeostatic control mechanisms. We find in numerical simulations of recurrent networks with a realistic triplet-based spike-timing-dependent plasticity rule (triplet STDP) that homeostasis has to detect rate changes on a timescale of seconds to minutes to keep the activity stable. We confirm this result in a generic mean-field formulation of network activity and homeostatic plasticity. Our results strongly suggest the existence of a homeostatic regulatory mechanism that reacts to firing rate changes on the order of seconds to minutes.

Public Library of Science2013

Spike-timing dependent plasticity

Wulfram Gerstner

Spike Timing Dependent Plasticity (STDP) is a temporally asymmetric form of Hebbian learning induced by tight temporal correlations between the spikes of pre- and postsynaptic neurons. As with other forms of synaptic plasticity, it is widely believed that it underlies learning and information storage in the brain, as well as the development and refinement of neuronal circuits during brain development (e.g. Bi and Poo, 2001; Sjöström et al., 2008). With STDP, repeated presynaptic spike arrival a few milliseconds before postsynaptic action potentials leads in many synapse types to long-term potentiation (LTP) of the synapses, whereas repeated spike arrival after postsynaptic spikes leads to long-term depression (LTD) of the same synapse. The change of the synapse plotted as a function of the relative timing of pre- and postsynaptic action potentials is called the STDP function or learning window and varies between synapse types. The rapid change of the STDP function with the relative timing of spikes suggests the possibility of temporal coding schemes on a millisecond time scale.

2010

Memory formation and recall in recurrent spiking neural networks

Friedemann Zenke

EPFL2014