Energy Efficient Federated Learning with Hyperdimensional Computing (HDC)

This document is an arXiv feed (2026-02-27) containing multiple systems and ML-infrastructure papers. Key items: (1) FL-HDC-DP — energy-efficient federated learning at the wireless edge combining hyperdimensional computing (HDC) with differential privacy and joint optimization of HDC dimension, transmit power, and CPU frequency; (2) "Engineered Simultaneity" — a proofs-based analysis showing the NBBO is frame-dependent, creating unavoidable windows exploited by high-frequency traders (authors estimate ≈$5B/year extracted), raising market-integrity concerns; (3) several parallel-file-system/aut