This code-only package accompanies the manuscript “Federated Edge Learning over Optical-UAV Access Links With Deterministic-Payload Compression and Packetized Scheduling.”The software implements static mixed-resolution compression with error feedback (SMR-EF) and PacketFair scheduling for simulated federated learning over optical links between clients and unmanned aerial vehicles. SMR-EF represents model updates using a base packet and progressive enhancement packets. PacketFair assigns delivery levels subject to bandwidth, power, communication-time, and access-deadline constraints. Error feedback is computed from the reconstruction actually delivered to the server.The implementation supports investigation of communication efficiency, reconstruction quality, and scheduling feasibility within the CIFAR-10/ResNet-20 experimental setting described in the manuscript. This deposit provides source code and accompanying usage documentation for inspecting the implementation, reusing its components, and running new experiments.
This deposit provides the software and reproducibility evidence for Precision and Coverage in Robust Federated Learning. It includes compressed client-update codecs, robust aggregation, seven development candidates and a complete five-seed comparison of three methods. Archived checkpoints, predictions, confusion matrices, traffic logs and readable result tables support reviewer inspection. The selected method achieves 73.654% mean final test accuracy, with a sample standard deviation of 0.383 percentage points, under the specified CIFAR-10 simulation. Reviewers can verify saved evidence without retraining. Full training instructions are also provided. The evaluation covers one dataset and one specified attack configuration. After extraction, reviewers can run:python -m pip install -r requirements-review.txtpython review.py --output ../review_verification.json
This code-only package accompanies the manuscript “Federated Edge Learning over Optical-UAV Access Links With Deterministic-Payload Compression and Packetized Scheduling.”The software implements static mixed-resolution compression with error feedback (SMR-EF) and PacketFair scheduling for simulated federated learning over optical links between clients and unmanned aerial vehicles. SMR-EF represents model updates using a base packet and progressive enhancement packets. PacketFair assigns delivery levels subject to bandwidth, power, communication-time, and access-deadline constraints. Error feedback is computed from the reconstruction actually delivered to the server.The implementation supports investigation of communication efficiency, reconstruction quality, and scheduling feasibility within the CIFAR-10/ResNet-20 experimental setting described in the manuscript. This deposit provides source code and accompanying usage documentation for inspecting the implementation, reusing its components, and running new experiments.
This deposit provides the software and reproducibility evidence for Precision and Coverage in Robust Federated Learning. It includes compressed client-update codecs, robust aggregation, seven development candidates and a complete five-seed comparison of three methods. Archived checkpoints, predictions, confusion matrices, traffic logs and readable result tables support reviewer inspection. The selected method achieves 73.654% mean final test accuracy, with a sample standard deviation of 0.383 percentage points, under the specified CIFAR-10 simulation. Reviewers can verify saved evidence without retraining. Full training instructions are also provided. The evaluation covers one dataset and one specified attack configuration. After extraction, reviewers can run:python -m pip install -r requirements-review.txtpython review.py --output ../review_verification.json
This code-only package accompanies the manuscript “Federated Edge Learning over Optical-UAV Access Links With Deterministic-Payload Compression and Packetized Scheduling.”The software implements static mixed-resolution compression with error feedback (SMR-EF) and PacketFair scheduling for simulated federated learning over optical links between clients and unmanned aerial vehicles. SMR-EF represents model updates using a base packet and progressive enhancement packets. PacketFair assigns delivery levels subject to bandwidth, power, communication-time, and access-deadline constraints. Error feedback is computed from the reconstruction actually delivered to the server.The implementation supports investigation of communication efficiency, reconstruction quality, and scheduling feasibility within the CIFAR-10/ResNet-20 experimental setting described in the manuscript. This deposit provides source code and accompanying usage documentation for inspecting the implementation, reusing its components, and running new experiments.
Julian Hoxha· Figshare· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
