You are here

Fine-tuning and the stability of recurrent neural

Abstract A central criticism of standard theoretical approaches to constructing stable, recurrent model networks is that the synaptic connection weights need to be finely-tuned. This criticism is severe because proposed rules for learning these weights have been shown to have various limitations to their biological plausibility. Hence it is unlikely that such rules are used to continuously fine-tune the network in vivo. We describe a learning rule that is able to tune synaptic weights in a biologically plausible manner. We demonstrate and test this rule in the context of the oculomotor integrator, showing that only known neural signals are needed to tune the weights. We demonstrate that the rule appropriately accounts for a wide variety of experimental results, and is robust under several kinds of perturbation. Furthermore, we show that the rule is able to achieve stability as good as or better than that provided by the linearly optimal weights often used in recurrent models of the integrator. Finally, we discuss how this rule can be generalized to tune a wide variety of recurrent attractor networks, such as those found in head direction and path integration systems, suggesting that it may be used to tune a wide variety of stable neural systems.
Peer Reviewed: 
MacNeil, D., Eliasmith, C. (2011). Fine-tuning and the stability of recurrent neural networks. PLoS ONE.
Publication files: 


1. Unzip the OMS model (v1.4) which can be found at

2. Uncomment and modify the second and third lines of NIdemo.m such that the
file path points to the location where the OMS model was unzipped.

3. Run the NIdemo.m file in MATLAB

Two plots will be shown after the simulation is complete. The first is a plot
of the optimal, noisy and learned decoders. The second plot is an animation of
the integrator learning throughout the duration of the simulation.

* NOTE: The OMSv1_4_mod_NEF_NI.mdl file is based on a prior model from

Changes from the original model are described in a separate research paper.