ANC Workshop - Chenfei Ma, Patricia Rubisch

Tuesday, 14th November 2023

A More Intuitive and Interactive Human Machine Interface - Chenfei Ma

Abstract: Have you ever thought you can control robots/machine or electrical devices with a simple hand gesture? Myoelectric Pattern Recognition (MPR) can decode your hand movement, even intention,  only with several electromyography (EMG) sensors around your arm. However, research has focused mostly on pattern recognition (PR) algorithm, but not on human adaptation. Ideally,  users should be able to intuitively control the machine, as easily as controlling a mouse on a screen.  For the users to easily understand the algorithm and cooperate with it, we propose a new method to visualise the PR algorithms in real time. From the comparison experiment, a significantly better interaction performance has been obtained, as well as a longer retention of the controlling skill for the users. We believe the proposed method is sufficiently general to improve any human machine interaction.

Inhibition as a di-synaptic regulator of voltage-dependent synaptic plasticity - Patricia Rubisch

Abstract: Synaptic plasticity, the process by which synapses change in an activity-dependent manner, is assumed to be the basis of learning in the brain. Experimental evidence demonstrates that activity originating from other synapses in close proximity to an observed one can influence the outcome of plasticity. One of those di-synaptic effects is the influence of inhibitory activity. The degree by which inhibition can effect the outcome of plasticity events depends on the sensitivity to subthreshold fluctuations of the plasticity model. I demonstrate that in models with high sensitivity to the instantaneous membrane potential, fast and precise inhibitory spiking can regulate synaptic plasticity and increases competition between excitatory neurons. As a result, excitatory-inhibitory recurrent networks can perform input source separation and factorisation similar to Independent Component Analysis. Furthermore, naturalistic input leads to development of receptive fields indicating that the increased competition between synaptic weights leads to the development of functionally relevant structures. This work demonstrates that plasticity models sensitive to di-synaptic inhibitory influences on fast time sclaes are sufficient to learn nonlinear network computations and highlights the importance of the detailed study of di- and multisynaptic influences on plasticity.

Event type: Workshop

Date: Tuesday, 14th November

Time: 11:00

Location: G.03

Speaker(s): Chenfei Ma, Patricia Rubisch

Chair/Host: Yuelin Yao