Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Building Semantic Communication System via Molecules: An End-to-End Training Approach

Yukun Cheng, Wei Chen, Bo Ai

TL;DR

The paper addresses efficient information transmission for molecular communication by reframing the problem as task-driven inference. It introduces an end-to-end semantic JSCC framework that directly maps input data to molecule concentrations and a channel network to enable gradient-based training over a non-differentiable propagation channel. Experiments on CIFAR-10 show the semantic approach outperforms JPEG+LDPC baselines, particularly at low per-slot molecule budgets and under significant ISI, while the channel network enables effective training despite non-differentiability. This work advances molecular communications by integrating semantic objectives with differentiable training strategies for non-electrical channels, enabling robust, resource-efficient inference in constrained environments.

Abstract

The concept of semantic communication provides a novel approach for applications in scenarios with limited communication resources. In this paper, we propose an end-to-end (E2E) semantic molecular communication system, aiming to enhance the efficiency of molecular communication systems by reducing the transmitted information. Specifically, following the joint source channel coding paradigm, the network is designed to encode the task-relevant information into the concentration of the information molecules, which is robust to the degradation of the molecular communication channel. Furthermore, we propose a channel network to enable the E2E learning over the non-differentiable molecular channel. Experimental results demonstrate the superior performance of the semantic molecular communication system over the conventional methods in classification tasks.

Building Semantic Communication System via Molecules: An End-to-End Training Approach

TL;DR

The paper addresses efficient information transmission for molecular communication by reframing the problem as task-driven inference. It introduces an end-to-end semantic JSCC framework that directly maps input data to molecule concentrations and a channel network to enable gradient-based training over a non-differentiable propagation channel. Experiments on CIFAR-10 show the semantic approach outperforms JPEG+LDPC baselines, particularly at low per-slot molecule budgets and under significant ISI, while the channel network enables effective training despite non-differentiability. This work advances molecular communications by integrating semantic objectives with differentiable training strategies for non-electrical channels, enabling robust, resource-efficient inference in constrained environments.

Abstract

The concept of semantic communication provides a novel approach for applications in scenarios with limited communication resources. In this paper, we propose an end-to-end (E2E) semantic molecular communication system, aiming to enhance the efficiency of molecular communication systems by reducing the transmitted information. Specifically, following the joint source channel coding paradigm, the network is designed to encode the task-relevant information into the concentration of the information molecules, which is robust to the degradation of the molecular communication channel. Furthermore, we propose a channel network to enable the E2E learning over the non-differentiable molecular channel. Experimental results demonstrate the superior performance of the semantic molecular communication system over the conventional methods in classification tasks.
Paper Structure (14 sections, 12 equations, 9 figures, 5 tables)

This paper contains 14 sections, 12 equations, 9 figures, 5 tables.

Table of Contents

  1. Introduction
  2. Preliminary
  3. System Model
  4. Molecular propagation channel
  5. The Proposed molecular semantic communication system
  6. Network Architecture
  7. Channel Network
  8. Training Method
  9. Experiments
  10. validatoin of the proposed channel network
  11. Performance evaluation in classification tasks
  12. Experimental setup
  13. Simulation results
  14. conclusion

Figures (9)

  • Figure 1: The model of the point-to-point semantic molecular communication system.
  • Figure 2: Molecular propagation channel.
  • Figure 3: Architecture of the proposed system.
  • Figure 4: Architecture of the channel network
  • Figure 5: SIR values for sending two consecutive bits $"1"$ in a molecular propagation channel with $n_m$=10000, (a) Scenario 1 (Channel with a low flow velocity) (b) Scenario 2 (Channel with a high flow velocity).
  • ...and 4 more figures