Our Work

On CRAN. Accepted on 2026-09-25; the CRAN page and binaries follow as the mirrors update. install.packages("exametrika") A minor release. The estimates returned by LRA(method = "SOM&qu…

On CRAN. Accepted on 2026-09-25; the CRAN page and binaries follow as the mirrors update. install.packages("exametrika") A minor release. The estimates returned by LRA(method = "SOM") change (with an explicit seed, or with mic = TRUE / conf), so the fixes below could not ship as a patch. They came out of reading the R code against the Mathematica routine it was ported from (Module_LRA.wl). No other model's estimates change. Bug Fixes Ordinal biclustering rolled back only half of its state. When an EM sweep failed to improve the likelihood, Biclustering() on ordinal data restored the category reference array BCRM but left the boundary array BBRM at the rejected values. The two are then one iteration apart, and since test_log_lik is computed from BCRM while the shape-restricted parameter count is read off BBRM, the reported nparam, df, AIC, BIC and CAIC could describe a different iterate than the likelihood does. It affects method = "R" with estimation = "isotonic" only; the other branches derive the count from the filter matrix or the rank and field counts instead, and the binary implementation rolls back the array it later reads. Both arrays are now saved and restored together. In practice the rejected sweep differs from the accepted one by around 1e-10 by the time the branch is reached, so the counted number of distinct boundary levels -- and hence every figure above -- came out unchanged in the cases tested; the fix removes the inconsistency rather than a visible error. The SOM method reused one frozen presentation order whenever seed was given. somclus() reseeded inside the epoch loop with set.seed(seed), so every epoch presented the respondents in exactly the same order. Online learning relies on that order being redrawn: with it fixed, the bias of one particular ordering is never averaged out. The original Mathematica routine reseeds each epoch with SeedRandom[Total[uuu] + somt], and the seed = NULL path already matched it; only the user-supplied seed was affected. The epoch number is now added to the seed, so a run stays reproducible while the order changes from epoch to epoch. Estimates from method = "SOM" with an explicit seed change accordingly. mic = TRUE sorted the rank reference matrix once per epoch instead of once per respondent. The original applies Sort /@ refmat immediately after each respondent is presented, which changes the matrix that decides the next winner; deferring it to the end of the epoch is a different algorithm, not a cheaper form of the same one. The sort -- and the conf constraint alongside it -- now runs after every respondent. Estimates from method = "SOM" with mic = TRUE or conf change accordingly. Ties in the SOM winner search are resolved toward the larger rank, as in the original (Sort[Transpose[{mlrank, clsnum}]][[-1]]). which.max() had been picking the smaller rank. method = "SOM" no longer warns that it may not have converged. SOM runs an annealing schedule for maxiter epochs and has no convergence criterion -- neither does the original implementation -- so finishing the schedule is normal termination. The warning fired on every default run and converge was always FALSE, which left no way to tell a healthy run from a broken one. converge is now TRUE unless BIC.check early stopping was requested and failed to trigger within ten times maxiter. The messages on that path, and the documentation of BIC.check, now describe it as early stopping rather than a convergence test; its threshold is unchanged. LRA(method = "SOM") no longer overwrites the caller's random number stream. It reseeds once per epoch and never restored .Random.seed, so any simulation that called it silently lost its own stream. The state is now restored on exit. Performance The SOM inner loop moved to C++ (src/som_core.cpp). One epoch of online updates -- winner search, neighbourhood update, conf, mic and the prior update -- now runs in compiled code, while the presentation order is still drawn in R so that set.seed() keeps governing reproducibility. On J15S500 with nrank = 6 and maxiter = 1000, estimation went from 5.2 s to 1.9 s with mic = FALSE, and from 103.5 s to 2.6 s with mic = TRUE (the per-respondent sort that mic now requires would otherwise have made it twenty times slower). The rank posterior is also computed once after the schedule instead of every epoch, since only BIC.check reads it in between. Internal grconvertX(), grconvertY() and rasterImage(), used by the rasterised Array plot since 2.0.1, are now imported from graphics in NAMESPACE. This clears the R CMD check note about undeclared global functions. Test suite All 3,536 tests pass. SOM had no tests before this release; tests/testthat/test-lra-som.R now pins it against a reference implementation written in the test file, since SOM draws from a different random number engine than Mathematica and cannot be matched to it number for number. See NEWS.md for the full changelog.

These files are the replication package for an empirical study of LLM-based software design pattern detection from compiled Java bytecode. The models never see source code. A Soot-based pipeline extra…

These files are the replication package for an empirical study of LLM-based software design pattern detection from compiled Java bytecode. The models never see source code. A Soot-based pipeline extracts five anonymized representations from the bytecode of nine P-MARt systems: decompiled Java-like code (Decomp), class hierarchy (CH), call graph (CG), control-flow graph (CFG), and intermediate representation (IR). We combine these into four inputs (Decomp, CH+CG, CFG+IR, and All). For each of the 590 P-MARt-annotated focus classes (508 from seven development projects and 82 from the held-out QuickUML and DrJava), one prompt per GoF pattern family (creational, structural, and behavioral) is sent with each input, and the model predicts pattern participation and participant roles. The study covers three LLMs (GPT-5.4, Gemini 2.5 Flash, and DeepSeek v4 Flash), a controlled comparison of rule-guided and direct prompting with GPT-5.4, and four supervised embedding-based baselines (CodeBERT, CodeT5, CodeGPT-Multilingual, and RoBERTa) that are trained on the development projects and compared with the LLMs on the two held-out projects. The package contains the project JARs, the extraction, prompting, evaluation, and baseline scripts, the prompt templates, the anonymized representations given to the models, and every prompt and response from the study, so the reported results can be recomputed without new API calls. All identifiers in the model inputs are anonymized to limit recognition of the original projects. The README gives step-by-step replication instructions, and SHA256SUMS lists a checksum for every file.

Supplementary material for: Nguyen Van Trung, Truong Hai Bang, "When does fuzzy control make swarm search adaptive? A conditional analysis and controlled study of fuzzy-controlled PSO and GWO for text…

Supplementary material for: Nguyen Van Trung, Truong Hai Bang, "When does fuzzy control make swarm search adaptive? A conditional analysis and controlled study of fuzzy-controlled PSO and GWO for text feature selection" (submitted to Applied Soft Computing). The archive contains the code, configuration and run records behind every table and figure of the article: - src/: pooling and de-duplication of LIAR and ISOT, the ten stratified 80/10/10 partitions, the Mamdani controllers, AF-PSO, FW-FGWO, the binary PSO/GA/GWO wrappers, the continuous benchmarks, the statistical tests (corrected resampled t-test, Holm adjustment) and the scripts that generate all tables and figures.- data/: download script and SHA-256 checksums for LIAR and ISOT (the datasets are not redistributed).- records/: one JSON record per method, dataset and partition, with per-item test predictions, controller traces and feature masks; schema and a sample record included.- scripts/run_one_partition.py: re-runs every method on one partition from scratch (about 4 minutes on one CPU core) and checks that the per-item predictions match the released records.- environment.txt, requirements.txt: hardware and library versions used for the published runs. Version 1.1.0 adds the 20-draw random-subset controls and the primary record set used in the revised article. Code: MIT licence. Run records: CC BY 4.0.

This paper presents a low-cost embedded platform integrating photovoltaic (PV) energy monitoring, maximum power point tracking (MPPT), and Internet of Things (IoT) communication for real-time control …

This paper presents a low-cost embedded platform integrating photovoltaic (PV) energy monitoring, maximum power point tracking (MPPT), and Internet of Things (IoT) communication for real-time control and performance evaluation. The proposed system is built around an ESP32 microcontroller with dual INA219 sensors and a custom-designed DC–DC boost converter operating at 20 kHz. Two MPPT strategies, perturb and observe (P&O) and fuzzy logic control (FLC), are implemented and experimentally compared under identical outdoor conditions. Experimental results obtained under natural irradiance demonstrate that the FLC approach achieves faster convergence and reduced steady-state oscillations compared to P&O. Specifically, FLC improves tracking efficiency from 82.5% to 84%, reduces duty cycle ripple, and increases average extracted power by approximately 1.5%. The system also enables real-time acquisition of I–V and P–V characteristics through duty-cycle sweeping. The proposed architecture integrates sensing, control, and IoT telemetry using MQTT and a ThingsBoard dashboard, enabling continuous monitoring with negligible communication latency relative to the MPPT update period. This work contributes a reproducible, low-cost testbed that bridges embedded control, power electronics, and IoT-based monitoring, with potential applications in both education and rapid prototyping of smart PV systems.

This repository contains the data and source code accompanying Matsubayashi, H. T., Razavi, S., Tahara, Y. O., Akenuwa, O. H., Rock, T. W., Nakajima, D., Otsuka-Yamaguchi R., Nakamura, H., Kramer, D. …

This repository contains the data and source code accompanying Matsubayashi, H. T., Razavi, S., Tahara, Y. O., Akenuwa, O. H., Rock, T. W., Nakajima, D., Otsuka-Yamaguchi R., Nakamura, H., Kramer, D. A., Matsuura, T., Chen, B., Lee, C. T., Miyata, M., Murata, S., Nomura, S. M., Inoue, T. (2026). Light-guided actin polymerization drives directed motility in protocells. 

Simulation data and postprocessing code supporting the manuscript Mechanism-separated lattice and cavity markers for differential liquid refractive-index readout. A square array of silver nanocylinde…

Simulation data and postprocessing code supporting the manuscript Mechanism-separated lattice and cavity markers for differential liquid refractive-index readout. A square array of silver nanocylinders on an Ag/SiO2/Ag stack supports two reflection minima governed by two different optical length scales: the long-wave marker is pinned to the ambient diagonal Rayleigh diffraction threshold, and the short-wave marker is the second longitudinal resonance of the dielectric spacer. This deposit contains every input the manuscript's figure and number producers read, those producers, the derived analysis records, and the eight published figures. Contents. Ambient-index, spacer, upper-film, radius, height, angle and period sweeps; baseline field cuts at both marker wavelengths; pattern-removal and dielectric-cylinder controls; order-resolved complex-field coefficients and the height-resolved decay record; paired-coordinate metric tables; the analysis code and the manuscript sources needed to regenerate the figures and reported values. Reproducing. Requires Python 3.11+ with NumPy, SciPy and Matplotlib. Run verify_deposit.py to check all files against MANIFEST.sha256, then build_figures.py, build_strengthled_figure.py and analyze_sensing.py. Verified from a clean extraction: all three producers complete and reproduce the manuscript's reported values, with a maximum relative deviation of 1.9×10-7 arising from iterative fits — far below the precision printed in the manuscript. Scope. The study is numerical throughout; no experimental data are included and none were taken. Calculations used Ansys Lumerical FDTD 2024 R1 with the solver's tabulated Palik silver and silica. The proprietary solver project (.fsp) files are not redistributed — one exceeds 400 MB and they are not inputs to any producer here — but the numerical configuration recovered from them is included in open form as fsp_configuration.json. They are available from the corresponding author on reasonable request. Quantities come from three simulation series with distinct configurations and are not pooled; reported linewidths and Q values are descriptors of sampled reflection minima, not intrinsic or pole quantities, and no detection limit or measurement precision is implied.

Solar photovoltaic (PV) performance is significantly affected by surface temperature and panel cleanliness, both of which influence energy conversion efficiency. Elevated temperatures reduce power out…

Solar photovoltaic (PV) performance is significantly affected by surface temperature and panel cleanliness, both of which influence energy conversion efficiency. Elevated temperatures reduce power output, while dust accumulation decreases effective solar irradiance. This study proposes an Internet of Things (IoT)-based integrated cooling and cleaning system as its main contribution, enabling real-time monitoring, intelligent control, and remote operation within a unified framework. The system utilizes an NTC 10K temperature sensor for continuous monitoring, while an ESP8266 microcontroller functions as the central controller for data processing, wireless communication, and web-based interfacing. When the panel temperature exceeds a predefined threshold (40–42 °C), The IoT-enabled system automatically activates a water pump and wiper mechanism to simultaneously reduce temperature and remove surface contaminants. Experimental results show a significant temperature reduction from approximately 41–43 °C to 30–33 °C. This improvement leads to an increase in voltage output of 0.5–1.2% and current output of 3–6%, depending on load conditions. The cleaning mechanism further enhances irradiance absorption by maintaining panel surface clarity. These findings demonstrate that IoT integration enables adaptive and data-driven optimization of PV performance. The proposed system offers a practical and scalable engineering solution to improve efficiency, reliability, and long-term operation of PV installations in real-world applications.

Literature sources: http://dx.doi.org/10.4067/S0717-77122005000100007https://doi.org/10.2134/agronmonogr9.2.c59https://doi.org/10.5479/si.1943-667X.42https://doi.org/10.1111/j.1439-0469.1985.tb00581.x…

Literature sources: http://dx.doi.org/10.4067/S0717-77122005000100007https://doi.org/10.2134/agronmonogr9.2.c59https://doi.org/10.5479/si.1943-667X.42https://doi.org/10.1111/j.1439-0469.1985.tb00581.xhttps://doi.org/10.1093/sysbio/45.2.223https://doi.org/10.1007/s00435-021-00519-3https://doi.org/10.1016/B978-0-12-385028-7.00007-Xhttps://doi.org/10.1098/rspb.2003.2631https://doi.org/10.1007/s11430-010-4084-yhttps://doi.org/10.1111/j.1558-5646.2009.00908.xhttps://doi.org/10.1073/pnas.94.10.51http://hdl.handle.net/10088/3418https://doi.org/10.1111/pala.12376ISBN:9783110272536https://doi.org/10.1111/j.1439-0469.2008.00478.xhttps://doi.org/10.1016/j.jtbi.2007.01.008https://www.jstor.org/stable/2411987https://doi.org/10.1016/j.tree.2015.03.003ISBN:0643065717https://doi.org/10.1242/dev.141507https://doi.org/10.1111/j.1558-5646.1994.tb05300.xhttps://doi.org/10.1093/icb/40.3.340https://books.google.com/books?id=IURZDwAAQBAJhttps://doi.org/10.1002/dvg.22748https://doi.org/10.1111/j.1469-185X.1984.tb00411.xhttps://doi.org/10.1134/S0031030114120107https://www.researchgate.net/publication/7547741_The_convoluted_evolution_of_snail_chiralityhttps://doi.org/10.1007/s00114-05-0045-2https://doi.org/10.1016/B978-0-12-385026-3.00011-5

Abstract This study evaluates contemporary empirical calibrations of cosmic acceleration by re-examining thecritical progenitor-age dependencies reported in Type Ia supernova data through the determin…

Abstract This study evaluates contemporary empirical calibrations of cosmic acceleration by re-examining thecritical progenitor-age dependencies reported in Type Ia supernova data through the deterministic geo-metric framework of Tone Theory. In this Grand Unified Theory, dark energy is formally re-derived as anative mechanical consequence of global boundary constraints and surface tension equations, manifestingas a Newtonian third-law recoil driven by the localized accumulation of gravitational curvature within amanifold. As the expanding Boltzmann transport volume grows, the structural resistance of the materialsubstrate relaxes, driving an apparent outward acceleration across the peripheral boundaries of space.By analyzing the mathematical parameters of recent empirical supernova evaluations with geometric pre-cision, we demonstrate that alleged redshift-dependent luminosity variations are completely governed bythe physical mass-step and selection-bias boundary conditions of the continuous canvas mesh, reinforcingcosmic acceleration without requiring an uncaused cosmological constant. PACS Codes: 95.36.+x, 98.80.Es, 97.60.Bw, 47.10.A-Keywords: Tone Theory, Dark Energy, Type Ia Supernovae, Substrate Continuum Mechanics, GeometricBoundary ConditionsCopyright & DOI Metadata: © 2026 A. Santana. All rights reserved.DOI:10.5281/zenodo.22949322