OSCILLATORY NEURAL NETWORKS

About Us


Although the electrical activity of a single neuron seems like train of spikes, the activity of a population of neurons at various scales measured in terms of LFP, BOLD, fMRI or EEG signals looks like continuous and periodic oscillations of various frequency bands (delta, theta, gamma, alpha). The stable limit cycle attractor of nonlinear dynamical system can universally be adapted to model electrical activity from single cell level to the cumulative of population of neurons at various scales.

Objectives

  • Primary objective is to propose generalized oscillatory neural network model capable of function approximation, classification, predictive modelling and designing controllers.
  • Propose a model for primate auditory cortex depicting tonotopic map architecture.
  • As this kind of network is capable of Fourier decomposing any time series signal, exploring the scope of other signal processing capabilities like filtering.
  • Modelling an oscillatory feed forward neural network to design the controller for gaiting rhythm generation as the muscle activation measured in terms of EMG signal is nearly periodic in nature during gait.
  • Experimentally recorded time series signals like EEG/ EMG/ fMRI signal classification.

Glimpse

A Glimpse of our work

Crawl Rotate Right

Crawl Turn Right

Crawl Turn Left

Crawl Backward

Crawl Forward

Higher Level Controller

Lower Level Controller

Current Projects/Problems

Current research works and projects

Power Coupling

Investigating phase encoding capability of network of Hopf oscillators using special kind of coupling strategy called “power coupling”.

Encoding and Prediction using Hopf Oscillators

Exploring the various regimes of Hopf oscillators to solve problems like encoding or predictive modelling.

Initialization for retrieving time series signal

Addressing initialization problem while reconstructing/retrieving stored time series signal from a Hopf field like oscillatory memory network.

Oscillatory LISSOM and CNN model

Constructing oscillatory neural network model equivalent for LISSOM and CNN.



Publications

Some selected recent publications

View Paper

Redirect

An Oscillatory Neural Autoencoder Based on Frequency Modulation and Multiplexing.

Soman, K., Muralidharan, V., Chakravarthy, V. S.

Front. Comput. Neurosci. 2018.
doi: 10.3389/fncom.2018.00052

Our Team

Our team members

Contact Us

Feel free to reach out to us.

Computational Neuroscince Lab
IIT Madras, Chennai 600036

subhodeep555@gmail.com

+91 779 726 6960