Distributed Indirect Source Coding with Decoder Side Information

Jiancheng Tang,Qianqian Yang,Deniz Gündüz
2024-05-22
Abstract:This paper studies a variant of the rate-distortion problem motivated by task-oriented semantic communication and distributed learning problems, where $M$ correlated sources are independently encoded for a central decoder. The decoder has access to a correlated side information in addition to the messages received from the encoders, and aims to recover a latent random variable correlated with the sources observed by the encoders within a given distortion constraint rather than recovering the sources themselves. We provide bounds on the rate-distortion region for this scenario in general, and characterize the rate-distortion function exactly when the sources are conditionally independent given the side information.
Information Theory
What problem does this paper attempt to address?
### Problems Addressed by the Paper This paper studies a variant of the multi-terminal source coding problem, inspired by task-oriented semantic communication and distributed learning problems. Specifically, the paper considers the case of \( M \) correlated sources being independently encoded, with these sources received by a central decoder. Besides receiving messages sent by the encoders, the decoder can also access side information related to these sources. The decoder's goal is to recover a latent random variable \( T \) related to the sources observed by the encoders, under a given distortion constraint, rather than directly recovering the sources themselves. ### Main Contributions 1. **Bounds on the Rate-Distortion Region**: - The paper provides general bounds on the rate-distortion region for this scenario. - It precisely characterizes the rate-distortion function when the sources are conditionally independent given the side information. 2. **Theoretical Analysis**: - By introducing auxiliary random variables \( W_1, W_2, W_3 \), a feasible rate-distortion region \( R_a(D) \) is derived. - A general outer bound \( R_o(D) \) is derived, and it is shown that these two regions coincide when the sources are conditionally independent, thus fully determining the rate-distortion function. ### Application Background - **Task-Oriented Semantic Communication**: The decoder only needs to reconstruct task-oriented information implied by the sources, such as extracting hidden features from scenes captured from multiple different angles. - **Distributed Learning**: In federated learning, the server recovers the updated global model based on messages received from all clients. ### Key Techniques - **Rate-Distortion Theory**: Analyzing the multi-terminal source coding problem through rate-distortion theory. - **Auxiliary Random Variables**: Introducing auxiliary random variables \( W_1, W_2, W_3 \) to simplify the problem analysis. - **Conditional Independence**: Simplifying the expression of the rate-distortion function when the sources are conditionally independent given the side information. ### Conclusion Through rigorous mathematical derivation, the paper provides a complete characterization of the rate-distortion region for a multi-terminal source coding problem, particularly when the sources are conditionally independent. These results have significant theoretical and practical implications for fields such as task-oriented semantic communication and distributed learning.