Implements semantic equalization for DeepJSCC, addressing mismatched latent spaces between transmitter and receiver models. Evaluates linear, neural, and zero-shot equalizers for aligning heterogeneous semantics. Enables robust, efficient communication in AI-native wireless systems.
This repo implements a MIMO-based method for semantic communications, learning precoder/decoder pairs to compress latent spaces and align semantics across devices. Includes both a linear ADMM-based model and a neural model under power and complexity constraints.
This repository tackles latent space misalignment in multi-agent AI-native semantic communications. It introduces a federated approach where an access point shares a semantic encoder, while user devices use local semantic equalizers to enhance mutual understanding.
This repository contains the implementation of our over-the-air semantic alignment framework using stacked intelligent metasurfaces (SIMs), which aligns latent representations directly in the wave domain to overcome mismatch between heterogeneous semantic encoders without adding extra digital processing.
To associate your repository with the semantic-equalization topic, visit your repo's landing page and select "manage topics."