Efficient online update of transmitter parameters in molecular autoencoder systems

Develop an efficient method to update the parameters of the neural-network-based transmitter within autoencoder-based end-to-end molecular communication systems during deployment, ensuring adaptation to time-varying molecular channels while minimizing computational and energy costs.

Background

The survey highlights autoencoders as a promising end-to-end solution for molecular communication, jointly optimizing transmitter and receiver operation without requiring closed-form channel models. However, real molecular channels are time-varying, which necessitates continual adaptation of neural network parameters in practice. While decoder-side fine-tuning is feasible, efficient strategies for updating transmitter-side parameters during real deployments—subject to strict energy and computational constraints—are not yet established.

This open question targets robust, low-overhead mechanisms to maintain performance in changing biological environments, particularly relevant to Internet of Bio-Nano Things applications where in situ updates are constrained by hardware and biocompatibility limitations.