An arXiv preprint presents a unified Rényi framework for composite binary testing, combining finite-sample bounds with asymptotic results that separate two Type II error regimes.
A theoretical result forces the number of cells in a bent partition to be a prime power and narrows its exponent using the dimension of the underlying finite space.
A preprint by Junmin An and Jon-Lark Kim gives exact formulas for the shortest self-orthogonal and LCD embeddings of linear codes over the ring Fq + uFq.
A model-based system for wireless sensor networks produced semantically enriched measurements, while most micro-ontology generation took place within 500 microseconds. Compression depended on the serialization format, and the reported engineering results came from an illustrative robot-based setup.
2020 46th Euromicro Conference on Software Engineering and Advanced Applications (SEAA), Portoroz, Slovenia, 2020, pp. 561-5684 min read
A theoretical analysis finds that generalized covering codes with product-form centers reach the ordinary sphere-covering rate, while linear codes over prime-power alphabets reach the same asymptotic benchmark.
In simulations and controlled hardware experiments, MagPie completed all reported first rendezvous, while variable energy and imposed clock-loss stress reduced delivery in modeled networks.
A modeling study compares five distributed tree-search strategies, finding strong agreement for exact cases but large errors in one shortcut and an unresolved capacity-limited case.
A virtual test of a three-router campus-core design reported interconnectivity and shortest-path routing, while physical validation and link-failure testing remain for future work.
A proposed routing system for multi-hop UAV networks reached about 96% cumulative packet arrival by the deadline in simulation and nearly 98% afterward, ahead of three multipath baselines.
A simulation case study reports that CTS kept a UAV-mounted RIS network above 90% of an optimal data-rate benchmark and delivered orders-of-magnitude gains in energy efficiency over some strategies. The authors say the approach could help under tight onboard energy and processing limits, but prototypes and field trials are still needed.
A simulated communication system using a shared ontology sent less scene information and reported higher question-answering accuracy than three comparison methods.
A simulation study reports that an adaptive framework could support up to 50 simultaneous public-safety MCX users while limiting harm to ordinary mobile traffic, but the evidence comes entirely from modeled network conditions.
In modeled tests, STAR-CRDT reported a 29.3% higher episode return than TD3+BC, with higher communication and sensing measures and a 54.2% lower collision-risk count in random mobility.
FedRep recorded the highest reported weighted sum-rate in a simulated three-cell scenario, while the reported rankings varied by base station and training stage.
The study examines how to maximize the entropy of a grouped probability distribution without exceeding a set limit, comparing an exact dynamic program with two greedy approximations.
2026 IEEE International Symposium on Information Theory (ISIT), Guangzhou, China, 2026, pp. 1-64 min read
A cloud-management framework reported lower average power use, fewer active servers and higher simulated profit than OP-PMF− across Google Cluster Data and PlanetLab traces.
IEEE Transactions on Services Computing 20244 min read
Theoretical analysis finds explicit lower bounds linking excess quadratic ball discrepancy to fixed-radius covering defects across several perfect-code settings. It also gives special one- and two-error certificates, while leaving optimal constants and closeness to a particular code unresolved.
In a synthetic benchmark, spaPS reached around 250,000 messages per second with spatial-filter mappings in broker memory, versus around 150,000 with mappings in the database.
Simulation-based tests of REE-TM reported higher reliability and lower energy use than two reduced versions, while resource utilization was 1% to 4% below OPTIMAL.
IEEE Transactions on Cloud Computing 20253 min read
A qubit-based model reported lower prediction errors than several comparison methods across six benchmark workload traces, but the simulated study did not test live cloud operations.
IEEE Transactions on Systems, Man, and Cybernetics: Systems 20254 min read
A simulated satellite system translated natural-language missions into radio policies, while deterministic software handled the feasibility-critical configuration tasks.
A new arXiv preprint shows that redesigning finite spherical codebooks as signal-to-noise ratio changes can reduce the leading constant in high-SNR error probabilities. The result is exact for simplex codebooks and the two orthoplex endpoints, but remains conjectural across the intermediate range.
A Docker-based 5G standalone testbed found that OAI and srsRAN had broadly similar radio-connection setup times, while srsRAN was generally steadier in throughput, video delivery, live streaming, and cloud gaming under the tested conditions.
An arXiv preprint reports that SlimTCP was fastest in selected data-centre benchmarks, but its authors say the tests do not settle whether simplification itself explains the result.
A history-aware recurrent neural network was tested as a way to manage changing workloads across GPUs in a three-dimensional multicellular simulation. The reported results included more even computation and lower migration costs, while the modeled epidermal sequence was not a separate biological validation study.
A proof-based study of a standard bivariate Gaussian model confirms both Yu–Tan conjectures and identifies a strictly larger exact common-information value for positive correlation.
A discrete-event simulation reported near-zero expected total cost for Slasher at a 15% power-reduction goal, while CPU-frequency tests found median retained capacity of 48% at 40% lower dynamic power.
A preprint describes psRL, a system for distributed agentic-RL training that reuses repeated input prefixes. It reports higher throughput, lower memory use and stronger scaling than veRL-based comparisons, while run-to-run variability and statistical tests were not reported.
A modeling study tests ShardMeter against measured operation and runtime data, then simulates a decentralized case in which 99% of maximum throughput corresponds to 21.69% lower cost.
A preprint describes BotScan, a replay-based system that prioritizes IP segments and ports to search for live IoT botnet command servers. In the reported tests, it found more servers than a baseline at the same probing budget and used fewer computing resources, although the evaluations covered selected datasets and address spaces.
A conceptual preprint defines wireless resilience as the system-level capacity to maintain and restore service performance in dynamic and uncertain environments. It organizes the idea into four dimensions, proposes six metrics and uses physical links and unmanned aerial vehicle networks as illustrative cases rather than empirical groups.
An arXiv preprint reports that a depth-averaged layered simplex estimator stayed competitive with classical probability estimators on large-alphabet tests, but did not dominate across all target distributions.
Tests of ParaWeb found large gains on compute-heavy numeric and image workloads, but performance varied sharply by programming pattern, runtime and implementation variant.
An arXiv preprint describes a real interference-alignment method for an active intelligent reflecting surface and reports a higher simulated sum rate with about a 260-fold speedup over WMMSE. The evidence is limited to a modeled system with synthetic channels and does not establish performance in real networks.
An arXiv preprint reports lower task-weighted radio-model error and severe-load backlog with RMWorld in simulated tests. The advantage was not universal, came with higher computational cost, and has not been tested on measured channels.