Task-Oriented Low-Label Semantic Communication With Self-Supervised Learning

2025-05-26

AI Summary

To address the limited task inference performance of task-oriented semantic communication under low-labeling regimes, this paper proposes SLSCom, a self-supervised semantic communication framework. Methodologically, SLSCom integrates self-supervised classification and reconstruction, contrastive feature learning, information bottleneck modeling, and end-to-end source–channel joint optimization. It is the first framework to jointly incorporate contrastive self-supervised learning and the information bottleneck principle into semantic encoding, enabling task-oriented unsupervised pretraining and zero- or few-shot downstream adaptation. Evaluated on image classification over multipath wireless channels, SLSCom significantly outperforms conventional digital communication and state-of-the-art deep learning–based semantic communication methods. Crucially, it maintains robust high performance even with minimal labeled data or when unlabeled data are task-irrelevant, effectively alleviating the labeling dependency bottleneck.

Technology Category

• Machine Learning: Semi-Supervised Learning
• Search and Optimization: Learning to Search
• Natural Language Processing: Sentence-level Semantics, Textual Inference, etc.

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Economics, Online Markets and Human Computation: Humans versus LLMs for data annotation and labeling

Abstract

Task-oriented semantic communication enhances transmission efficiency by conveying semantic information rather than exact messages. Deep learning (DL)-based semantic communication can effectively cultivate the essential semantic knowledge for semantic extraction, transmission, and interpretation by leveraging massive labeled samples for downstream task training. In this paper, we propose a self-supervised learning-based semantic communication framework (SLSCom) to enhance task inference performance, particularly in scenarios with limited access to labeled samples. Specifically, we develop a task-relevant semantic encoder using unlabeled samples, which can be collected by devices in real-world edge networks. To facilitate task-relevant semantic extraction, we introduce self-supervision for learning contrastive features and formulate the information bottleneck (IB) problem to balance the tradeoff between the informativeness of the extracted features and task inference performance. Given the computational challenges of the IB problem, we devise a practical and effective solution by employing self-supervised classification and reconstruction pretext tasks. We further propose efficient joint training methods to enhance end-to-end inference accuracy over wireless channels, even with few labeled samples. We evaluate the proposed framework on image classification tasks over multipath wireless channels. Extensive simulation results demonstrate that SLSCom significantly outperforms conventional digital coding methods and existing DL-based approaches across varying labeled data set sizes and SNR conditions, even when the unlabeled samples are irrelevant to the downstream tasks.

Problem

• Enhancing task inference with limited labeled samples
• Self-supervised semantic communication for efficient feature extraction
• Balancing feature informativeness and task performance via IB

Innovation

• Self-supervised learning for semantic communication
• Task-relevant semantic encoder with unlabeled samples
• Information bottleneck for feature-task balance