In-Context Learning for Deep Joint Source-Channel Coding Over MIMO Channels

Meng Hua, Wenjing Zhang, Chenghong Bian, Deniz Gündüz · 2025

Read In-Context Learning for Deep Joint Source-Channel Coding Over MIMO Channels by Meng Hua, Wenjing Zhang, Chenghong Bian, Deniz Gündüz (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2512-01567.txt.

Description

Meng Hua and co-authors blend in-context learning with deep joint coding for image sends. Joint source-channel coding sends images over noisy links in one learned step. Over MIMO links, with many antennas, the task gets harder. Channel states shift fast.

Large models can learn from a few examples in context. The team uses that skill to adapt the coder on the fly. Few-shot adaptation avoids retraining for each new channel. Adaptive coding could keep image quality high on mobile links where conditions change by the second. Context examples ride in the signal itself, so the receiver needs no side channel to learn the sender's code for each new image.

Catalog record

CREATORMeng Hua, Wenjing Zhang, Chenghong Bian, Deniz Gündüz
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE64 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2512-01567

Browse the full library · About the Tower of Babel