Quantum
Encryptability As a Coordinate Choice: Depth-One Homomorphic Federated Learning of Quantum Neural Networks
Publication: quant-ph updates on arXiv.org
Author: Marcel Mordarski, Nathan Mani, Arshad Patel, William Knottenbelt, Roberto Bondesan
Source published: 28/09/2026, 04:00:00 UTC
Briefing created: 28/09/2026, 04:20:28 UTC
Based on a feed excerpt; full article content has not been acquired. Source evidence retrieved 28/09/2026, 04:13:25 UTC.
Briefings
A new protocol enables encrypted federated learning of quantum neural networks by exploiting the bilinear nature of unit-quaternion updates, which eliminates the prohibitive computational costs associated with traditional coordinate choices. The study confirms zero measurable utility tax in a five-seed experiment and demonstrates that encryption noise does not regularize convergence, achieving high fidelity on a 156-qubit processor.
Key points
- The protocol uses unit-quaternion coordinates to represent variational quantum circuit weights, allowing encrypted updates to remain at a single multiplicative level instead of requiring prohibitive transcendental operations.
- A paired five-seed study confirmed zero measurable utility tax with an MSE difference of +9×10^-6 and a p-value of 0.92, while a noise-budget ablation falsified the hypothesis that encryption noise regularizes training.
Context and limitations
Results are limited to two cryptographic backends and a 156-qubit processor; the study covers only encrypted rotation updates and does not address other coordinate choices or hardware-specific noise sources.
Generated by Bittrees-hosted qwen35-2b sha256:aaf42c8b7c3c on 28/09/2026, 04:20:28 UTC. Based on publisher-provided feed evidence; may not cover the full original article. Check the source for details.
View archived briefing on IPFS
bafybeien6ne3imz4tq5k35h3s2vheb2fzclpnwvmhurlq4waowupa375be