UCL Division of Biosciences


NPP Seminar: Professor Claudia Clopath, Imperial College London

09 November 2022, 1:00 pm–2:00 pm

Claudia Clopath


Event Information

Open to

UCL staff | UCL students | UCL alumni




Charlette Bent-Gayle


G46 H O Schild Pharmacology LT
Medical Sciences and Anatomy
Gower Street
United Kingdom

Academic Host: Andrew MacAskill

Abstract: Experimental and computational studies suggest that motor cortex acts as a feedback controller, allowing for ‘on-the-fly’ movement corrections in response to afferent sensory feedback. However, it remains unclear whether feedback control relates to longer-term learning, and how this would be implemented in neural circuitry. Here, we tackled these questions by testing how a recurrent neural network (RNN) can use feedback to control its own output, and whether this process can enable learning. We built an RNN that received feedback signaling the error between its intended and observed output. An initial training phase that required producing a broad range of outputs (i.e., ‘movements’) enabled the model to learn to use this feedback to correct its output on-the-fly. After constructing this RNN, we tested directly whether the feedback signal used for online output correction could enable learning by guiding synaptic plasticity in the recurrent connections within the network. We devised a biologically plausible plasticity rule where the recurrent weight changes were proportional to the error feedback signals received by the postsynaptic neurons. This simple rule allowed the network to adapt to persistent perturbations (e.g., a ‘visuomotor rotation’) by changing its initial output pattern, a process that was mediated through recurrent connectivity changes. Remarkably, the model learned in a way that was similar to adaptation studies in humans: i) learning generalized to non-learned but similar movements and ii) followed multiple learning timescales. When we examined the network activity before and after adaptation, we found a signature of our learning rule that was also present in neural population recordings from monkey motor cortex. In short, this work links algorithmic models of motor control and learning to a biologically plausible implementation in neural circuitry, thus offering the potential to guide future experimental studies on the neural basis of motor learning.



About the Speaker

Claudia Clopath

Professor of Computational Neuroscience at Imperial College London

Professor Claudia Clopath is based in the Bioengineering Department at Imperial College London. She is heading the Computational Neuroscience Laboratory.

Her research interests are in the field of neuroscience, especially insofar as it addresses the questions of learning and memory. She uses mathematical and computational tools to model synaptic plasticity, and to study its functional implications in artificial neural networks.

Prof. Clopath holds an MSc in Physics from the EPFL and did her PhD in Computer Science under Wulfram Gerstner. Before joining Imperial College, she did postdoctoral fellowships in neuroscience with Nicolas Brunel at Paris Descartes and in the Center for Theoretical Neuroscience at Columbia University. She published highly cited articles in top journals such as Science and Nature, has given dozens of invited talks and keynotes around the world, and received various prizes such as the Google Faculty Award in 2015.

More about Claudia Clopath