Our Work

This repository provides a comprehensive implementation for class-conditional forecast model allocation in retail inventory replenishment, based on empirical research from a hardware distribution port…

This repository provides a comprehensive implementation for class-conditional forecast model allocation in retail inventory replenishment, based on empirical research from a hardware distribution portfolio (FERRETODO). The framework integrates ABC inventory segmentation with predictive model routing, evaluating six forecasting algorithms across aggregated demand classes ( A, B, and C ) derived from 35,399 transactional records covering 1,633 SKUs. The repository is designed to support retail logistics, supply chain management, demand forecasting, and inventory replenishment planning, providing a reproducible workflow to determine where advanced forecasting models yield operational value and where simple benchmarks suffice. Key components include: Transactional data cleaning, median unit price imputation, and weekly aggregation ABC inventory segmentation based on cumulative annual revenue thresholds (80% / 15% / 5%) Feature engineering (autoregressive lags 1–8, rolling statistics, calendar dynamics, and national holiday/high-season exogenous indicators) Statistical time-series modeling (ARIMA, SARIMA, ARIMAX with calendar regressors) Non-parametric machine learning and deep learning architectures (Random Forest, LightGBM, LSTM) Out-of-sample evaluation on holdout windows and expanding rolling-origin validation (12 decision points) Accuracy assessment using MAE, RMSE, MAPE, MASE, and RMSSE against Naïve and Seasonal Naïve benchmarks Statistical significance validation using the Diebold-Mariano test (with Harvey-Leybourne-Newbold and Newey-West corrections) Sensitivity analysis of model allocation across alternative ABC revenue cut-offs Reproducible research workflows, multi-seed LSTM stability checks, and high-resolution diagnostic plots

Mazkur ilmiy maqolada oliy ta'lim tizimida talabalarning fuqarolik faollik ko'nikmalarini shakllantirishda hamda ularning ilmiy salohiyatini oshirishda mavjud tarixiy, ijtimoiy-siyosiy hamda pedagogik…

Mazkur ilmiy maqolada oliy ta'lim tizimida talabalarning fuqarolik faollik ko'nikmalarini shakllantirishda hamda ularning ilmiy salohiyatini oshirishda mavjud tarixiy, ijtimoiy-siyosiy hamda pedagogik-psixologik muammolarni jahon va mahalliy tajribalardan kelib chiqib, bugungi holatini tahlil qilish orqali ilmiy-nazariy ma'lumotlarni amalda qo'llashni takomillashtirish bo'yicha xulosalar va takliflar ishlab chiqilgan. Jahon fanida fuqarolik faollikni ta'minlashda motiv va ehtiyojlarni o'rganishga yo'naltirilgan bir qator aqliy markaz (think tanks) va ilmiy-tadqiqot institutlarida izlanishlar olib borilmoqda. Ular qatoriga Amnesty International (Buyuk Britaniya), Strategik va innovatsion tadqiqotlar markazi (AQSH), Fridrix Ebert jamg'armasi – FES (Germaniya), Ijtimoiy-iqtisodiy tadqiqotlar markazi – CASE (Polsha), Fuqarolik jamiyati va notijorat sektori (Rossiya Federatsiyasi) kabilarning tadqiqotlarida inert (ing. inert – sust, harakatsiz, nofaol) holatdan faol fuqarolikka o'tish siyosiy, iqtisodiy, huquqiy bilimdonlikni talab etishi bilan birga, faol fuqarolik fenomeni jamiyatda ehtiyoj tarzida ifodalanishi bilan aloqadorligini ko'rsatgan. Markaziy Osiyoda fuqarolik jamiyatini rivojlantirishga oid majmuaviy tadqiqotlarda yoshlarda fuqarolik kompetensiyalarini shakllantirish doirasida yangi bilim va tushunchalarga ega bo'lish, xalqaro miqyosda maqsadli tadqiqotlarni hamkorlikda o'tkazish strategiyasini ishlab chiqish belgilangan. Unga muvofiq ta'lim muassasalarida huquqiy va iqtisodiy bilimlarni zamonaviy pedagogik texnologiyalar asosida olib borish, pedagoglarni tayyorlash fuqarolik faollikni ta'minlashning asosiy yo'llaridan ekanligi, mamlakatning ijtimoiy va iqtisodiy rivojlanishini ta'minlashning ma'rifiy asosi hisoblanishi belgilangan.

Mazkur ilmiy maqolada oliy ta'lim tizimida talabalarning fuqarolik faollik ko'nikmalarini shakllantirishda hamda ularning ilmiy salohiyatini oshirishda mavjud tarixiy, ijtimoiy-siyosiy hamda pedagogik…

Mazkur ilmiy maqolada oliy ta'lim tizimida talabalarning fuqarolik faollik ko'nikmalarini shakllantirishda hamda ularning ilmiy salohiyatini oshirishda mavjud tarixiy, ijtimoiy-siyosiy hamda pedagogik-psixologik muammolarni jahon va mahalliy tajribalardan kelib chiqib, bugungi holatini tahlil qilish orqali ilmiy-nazariy ma'lumotlarni amalda qo'llashni takomillashtirish bo'yicha xulosalar va takliflar ishlab chiqilgan. Jahon fanida fuqarolik faollikni ta'minlashda motiv va ehtiyojlarni o'rganishga yo'naltirilgan bir qator aqliy markaz (think tanks) va ilmiy-tadqiqot institutlarida izlanishlar olib borilmoqda. Ular qatoriga Amnesty International (Buyuk Britaniya), Strategik va innovatsion tadqiqotlar markazi (AQSH), Fridrix Ebert jamg'armasi – FES (Germaniya), Ijtimoiy-iqtisodiy tadqiqotlar markazi – CASE (Polsha), Fuqarolik jamiyati va notijorat sektori (Rossiya Federatsiyasi) kabilarning tadqiqotlarida inert (ing. inert – sust, harakatsiz, nofaol) holatdan faol fuqarolikka o'tish siyosiy, iqtisodiy, huquqiy bilimdonlikni talab etishi bilan birga, faol fuqarolik fenomeni jamiyatda ehtiyoj tarzida ifodalanishi bilan aloqadorligini ko'rsatgan. Markaziy Osiyoda fuqarolik jamiyatini rivojlantirishga oid majmuaviy tadqiqotlarda yoshlarda fuqarolik kompetensiyalarini shakllantirish doirasida yangi bilim va tushunchalarga ega bo'lish, xalqaro miqyosda maqsadli tadqiqotlarni hamkorlikda o'tkazish strategiyasini ishlab chiqish belgilangan. Unga muvofiq ta'lim muassasalarida huquqiy va iqtisodiy bilimlarni zamonaviy pedagogik texnologiyalar asosida olib borish, pedagoglarni tayyorlash fuqarolik faollikni ta'minlashning asosiy yo'llaridan ekanligi, mamlakatning ijtimoiy va iqtisodiy rivojlanishini ta'minlashning ma'rifiy asosi hisoblanishi belgilangan.

Supplementary repository accompanying the preprint 'Open-World Source-Free Adaptation for Underwater Acoustic Target Recognition under Naturally Contaminated Target Streams: Adaptation Survives, Post-…

Supplementary repository accompanying the preprint 'Open-World Source-Free Adaptation for Underwater Acoustic Target Recognition under Naturally Contaminated Target Streams: Adaptation Survives, Post-hoc Rejection Does Not' (to appear as Preprints 2026; preprint v1 (manuscript v1.3.3), September 2026). Fulfills the Data Availability statement: upon preprint posting, this archival repository releases the code, configuration files, split identifiers where permitted, trained-model metadata, and derived evaluation artifacts. Contents: all 17 analysis/training/evaluation scripts; per-run configs, histories, and summaries (3 directions x 5 methods x 3 seeds); aggregate metric JSON and per-seed CSVs; unknown-subgroup rescoring tables and cached per-segment logits; threshold-transfer cell data; feature-space scoring results with aggregate source-stage statistics; calibration/bootstrap/sklearn cross-check tables; FPR@80/90/95TPR operating-point sensitivity tables; figure sources; requirements.txt with dependency/environment notes; sha256 manifest. No raw audio is redistributed; model checkpoints and third-party cached embeddings are excluded per the underlying dataset licenses and are exactly reproducible with the deposited scripts. VTUAD is available under concept DOI 10.5281/zenodo.22274549; Oceanship/QiandaoEar22/ShipsEar from their respective providers under their stated licenses.

This article analyzes the theoretical and practical significance of improving the accounting of cash and cash equivalents in the context of the rapidly developing digital economy, based on Internation…

This article analyzes the theoretical and practical significance of improving the accounting of cash and cash equivalents in the context of the rapidly developing digital economy, based on International Financial Reporting Standards (IFRS / IAS 7 "Statement of Cash Flows"). The study highlights the challenges of recognizing, measuring, and reporting cash equivalents considering digital financial technologies (e-wallets, digital assets, online payment systems). The author provides scientific and practical recommendations aimed at enhancing cash flow transparency, reducing financial risks, and improving the investment attractiveness of enterprises.

Analysis code for "Multimodal characterisation of three infrared lasers for ambient mass spectrometry imaging": consensus ion list construction, crater-area and signal-per-crater-volume statistics, th…

Analysis code for "Multimodal characterisation of three infrared lasers for ambient mass spectrometry imaging": consensus ion list construction, crater-area and signal-per-crater-volume statistics, the chloride-adduct isotope envelope, the homogenate spectral histogram, and HDImaging-to-imzML conversion.

Mazkur maqolada maxsus ta'lim ehtiyojiga ega o'quvchilarning ijtimoiy-psixologik rivojlanishini qo'llab-quvvatlashda sun'iy intellekt texnologiyalarining imkoniyatlari tahlil qilinadi. Tadqiqotda mulo…

Mazkur maqolada maxsus ta'lim ehtiyojiga ega o'quvchilarning ijtimoiy-psixologik rivojlanishini qo'llab-quvvatlashda sun'iy intellekt texnologiyalarining imkoniyatlari tahlil qilinadi. Tadqiqotda muloqot va nutqni qo'llab-quvvatlovchi tizimlar, emotsiyalarni aniqlovchi affektiv hisoblash texnologiyalari, ijtimoiy robotlar yordamida terapiya, virtual reallik va adaptiv ta'lim platformalari kabi yo'nalishlar ko'rib chiqiladi. Tadqiqot natijalari sun'iy intellekt texnologiyalarining maxsus ta'lim ehtiyojiga ega o'quvchilarda ijtimoiy ko'nikmalar, emotsional o'zini o'zi tartibga solish va muloqot qobiliyatlarini rivojlantirishda sezilarli salohiyatga ega ekanligini, ammo ularni inson mutaxassisi ishtirokisiz emas, balki uni to'ldiruvchi vosita sifatida qo'llash zarurligini ko'rsatadi.

Deep learning-based analysis reveals patient-level proton radiation therapy trajectories using single-cell PBMC chromatin images

Supplementary repository accompanying the preprint 'General-Purpose Audio Foundation Models Improve Zero-Adaptation Cross-Water Transfer in Underwater Acoustic Target Recognition' (to appear as Prepri…

Supplementary repository accompanying the preprint 'General-Purpose Audio Foundation Models Improve Zero-Adaptation Cross-Water Transfer in Underwater Acoustic Target Recognition' (to appear as Preprints 2026; preprint v1.3, September 2026). This archive fulfills the preprint's Data Availability statement: upon preprint release it deposits the full evaluation and analysis code; 126 per-cell metric JSONs (within-corpus probes, 72-pair cross-corpus zero-shot transfer, VTUAD v1.2.0-label re-probes); aggregate bootstrap and protocol-audit JSONs; the o2q permutation-control numerics (5,000 runs per granularity: full F1 distributions, permuted label matrices, labels and recording ids); per-segment predictions for all 12 cross-transfer directions (FM probes and 3-seed + oracle ResNet-18 anchors); the Appendix B 12 x 6 matrix; the four-corpus audio index (258,893 segments, basenames only); figure sources; and the embedding caches of the series' own VTUAD corpus for all six models (uploaded as six per-model zips alongside the main archive). No raw audio is redistributed. Cached embeddings of the third-party corpora (ShipsEar, Oceanship, QiandaoEar22) are deliberately excluded under the data providers' terms (no redistribution of the data or derivative copies) and are exactly regenerable with the deposited scripts; every reported statistic can be verified from the deposited numeric artifacts. VTUAD itself: Zenodo concept DOI 10.5281/zenodo.22274549. A sha256 manifest (SHA256SUMS.txt) covers every file in the main archive. Version 2 supplement (2026-09-18): adds (a) the companion UDA benchmark's v2 seed-expansion target-oracle summary (aggregate statistics and per-seed results for all 12 directions: 10 seeds for the Oceanship-QiandaoEar22 two-domain matrix, 8 seeds for the seven VTUAD/ShipsEar-extension directions covered by the expansion, 3 original seeds for the three uncovered directions), which backs the target-oracle column of Table 4.3 of the preprint; (b) an environment file (requirements.txt); (c) a split-manifest and data-version note (per-corpus train/val/test segment counts, VTUAD v1.2.0 label-correction and QiandaoEar22 filename-collision documentation). All previously deposited artifacts are unchanged; the reported statistics are identical. Version 3 supplement (2026-09-22): adds the A10 strong-baseline round artifacts for the eight-model revision of the preprint - (a) per-cell probe metric JSONs, per-segment cross-corpus predictions, and the as-run extraction/analysis scripts for the two added encoders PANNs-CNN14 (official 32 kHz checkpoint, md5-verified) and BEATs (pretrained iter3+ AS2M checkpoint); (b) the eight-model analysis aggregate (protocol A/B audit, per-seed paired bootstrap, recording-cluster bootstrap) backing Tables 5/7/8 of the v2 preprint; (c) the 12-direction x 8-model appendix-B matrix (supersedes the six-model matrix in the v1 main zip); (d) VTUAD embedding caches of the two added encoders as separate per-model archives. All previously deposited artifacts are unchanged. Version 3.1 corrections (2026-09-29): the Protocol B selection rule in the analysis scripts was corrected to source-validation selection (8-model wins 7 of 12, 6-model 5 of 12; the 96 cross-transfer cell values, the oracle envelope, and Protocol A are unchanged), a train_class_counts logging bug was fixed in all 240 cross-transfer JSONs (accuracy/F1 values untouched), decode-failure handling was hardened (wav_decode_failures.json + P8_FAIL_FAST), SHA256SUMS regenerated, and per-package CORRECTIONS_20260929.md notes document every change. Two archives rebuilt: paper8_supplementary_repository_v2.zip and paper8_a10_supplement_v3.zip; all embedding caches and the UDA oracle summary are byte-identical to v3.

Supplementary repository accompanying the preprint 'Cross-Dataset Generalization of Underwater Acoustic Ship Recognition: A Four-Dataset Benchmark of Label and Acoustic Domain Shifts' (to appear as Pr…

Supplementary repository accompanying the preprint 'Cross-Dataset Generalization of Underwater Acoustic Ship Recognition: A Four-Dataset Benchmark of Label and Acoustic Domain Shifts' (to appear as Preprints 2026, 202609.0919). This archive fulfills the preprint's Data Availability statement: it deposits the frozen recording/clip-level train/validation/test splits (*_v1.json) for Oceanship, QiandaoEar22, VTUAD (pre-v1.2.0 experimental snapshot), and ShipsEar; the four-corpus class-mapping table with per-class exclusion rationales; the full evaluation and preprocessing pipeline code (make_splits, preprocess_audio, train_baseline, cross_library_eval, cross-dataset probes, leakage audit, recording-level cluster bootstrap); the trained ResNet-18 checkpoints used by the benchmark tables; per-seed and per-direction result artifacts including the recording-level bootstrap JSONs; and the segment manifests recording the VTUAD pre-v1.2.0 snapshot row by row (features/vtuad/manifest_{train,val,test}.csv; 133,784 extraction rows, 132,273 unique segments after content deduplication). No raw audio is redistributed: ShipsEar audio remains subject to its authors' educational-use agreement, and Oceanship / QiandaoEar22 / Ocean Networks Canada audio are available from their respective public releases. All frozen splits released here are reused in full by the companion papers of this series, so any experiment built on these splits is directly comparable with this benchmark.