Popular repositories Loading
-
DJSCC-Semantic-Equalization
DJSCC-Semantic-Equalization PublicImplements semantic equalization for DeepJSCC, addressing mismatched latent spaces between transmitter and receiver models. Evaluates linear, neural, and zero-shot equalizers for aligning heterogen…
-
semantic-alignment-mimo
semantic-alignment-mimo PublicThis 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…
-
multi-agent-semantic-alignment
multi-agent-semantic-alignment PublicThis 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 de…
-
topological-dictionary-learning
topological-dictionary-learning PublicOffical repository for the paper "Topological Dictionary Learning"
-
semantic-dict-sheaf
semantic-dict-sheaf PublicThis repository implements a dictionary-driven, sheaf-based semantic communication framework for heterogeneous AI agents, learning both the communication topology and sheaf maps for latent-space al…
Repositories
- rkhs-alignment Public
A closed-form kernel framework for semantic alignment of latent spaces, pairing a geometry-preserving Procrustes transformation with an orthogonality-constrained RKHS residual to capture nonlinear latent mismatch from a small set of kernel-herded semantic pilots.
- sheaf-based-federated-representation-learning Public
This repository contains the implementation of Sheaf-based Federated Representation Learning (SFRL), a decentralized framework where heterogeneous clients train together by exchanging only latent representations of shared pilot samples and align them through learnable sheaf restriction maps instead of a shared global latent space.
- sheaf-theoretic-channel-charting Public
A sheaf-theoretic framework for distributed multi-site channel charting that jointly learns neural encoders and inter-site orthogonal transport maps to achieve coherent cross-site alignment of local representations.
- semasia-datasets Public
Large-scale dataset of latent embeddings from ~1700 pretrained vision models across 8 image benchmarks. Enables studying latent geometry, benchmarking semantic alignment methods, and regressing how pretraining strategies shape embedding structure.
- semantic-alignment-via-sim Public
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.
- multi-agent-semantic-alignment Public
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.
- spaicom-site Public
- semantic-dict-sheaf Public
This repository implements a dictionary-driven, sheaf-based semantic communication framework for heterogeneous AI agents, learning both the communication topology and sheaf maps for latent-space alignment, supported by a sparse dictionary–based semantic denoising and compression module that builds a shared global semantic space.
- spml-lab Public
This repository hosts the laboratory exercises and supporting materials for the Signal Processing for Machine Learning course within the Master’s program in Data Science at Sapienza University of Rome.
Top languages
Loading…
Most used topics
Loading…