Our Work

Мазкур мақолада коррупцияга қарши курашишда превентив фаолиятнинг илмий-назарий асослари, превенция тушунчасининг мазмун-моҳияти ҳамда унинг турлари таҳлил қилинган. Шунингдек, прокуратура органларини…

Мазкур мақолада коррупцияга қарши курашишда превентив фаолиятнинг илмий-назарий асослари, превенция тушунчасининг мазмун-моҳияти ҳамда унинг турлари таҳлил қилинган. Шунингдек, прокуратура органларининг коррупцияга қарши курашишдаги ўзига хос ўрни, прокурор назорати, ҳуқуқни муҳофаза қилувчи органлар фаолиятини мувофиқлаштириш, коррупциявий хавф-хатарларни аниқлаш ва уларни бартараф этишга қаратилган чораларнинг превентив хусусияти тадқиқ этилган. Тадқиқот натижасида прокуратура органларининг превентив фаолиятини янада такомиллаштириш билан боғлиқ муаммолар аниқланиб, коррупциявий хавфларни баҳолашни хавфга асосланган ёндашув асосида ташкил этиш, превентив фаолият самарадорлигини баҳолашнинг аниқ мезонларини жорий этиш бўйича таклифлар илгари сурилган.

Reproducible covariance-robust Gaussian disclosure experiments, including shared-predictor semidefinite certificates, synthetic uncertainty families, and chronological electricity-data prediction diag…

Reproducible covariance-robust Gaussian disclosure experiments, including shared-predictor semidefinite certificates, synthetic uncertainty families, and chronological electricity-data prediction diagnostics.

Task-based multithreading and parallel reductions (milestone v3.13.0): CpuThreaded now threads inner products, warmed assembly refills, the stencil and vector-calculus operators and broadcast assignme…

Task-based multithreading and parallel reductions (milestone v3.13.0): CpuThreaded now threads inner products, warmed assembly refills, the stencil and vector-calculus operators and broadcast assignment, and CpuPolyester batches the same paths, all without new dependencies and with results equal to CpuSerial bit for bit (reductions to rounding). This minor release has one breaking change, listed first. Breaking VectorElement integer multi-indices follow AbstractVector rules (#348) On a 2D or 3D grid function, uₕ[i, j] no longer means grid point (i, j). It now follows AbstractVector's trailing-index semantics: uₕ[k, 1] == parent(uₕ)[k], and any other trailing index is a BoundsError. This is what generic code (sparse assembly, LinearAlgebra) already assumed. Index the grid with a CartesianIndex: uₕ[CartesianIndex(i, j)], or through reshape(uₕ), which returns a grid-shaped view of the same memory. Performance Threaded inner products without Polyester (#301) innerₕ, normₕ and the other reductions under CpuThreaded sum Threads.nthreads() static chunks in a fixed order, so a result is the same on every run. Masked reductions walk only the set bits, and separable weights are chunked along the last axis without per-point divrem. Measured crossover against CpuSerial on an Apple M2 at --threads=4: 100,000–300,000 elements. Threaded refills replay the recorded positions (#338) A warmed assemble! under Parallel() or CpuPolyester writes each entry to the nzval position recorded on the first assembly, instead of searching for it; assemble_add! does too. At 1M DOFs a threaded refill now takes 0.57–0.68 of the serial refill's time in 1D/2D/3D, where it used to be slower than serial in 2D and 3D. Threaded and batched operators (#356) The difference and average stencil engines (D₋ₓ! and its families, Mₓ!, jumps) and the vector-calculus operators (∇, div, curl, ε families) band along the last axis under CpuThreaded and CpuPolyester. Crossover against CpuSerial for D₋ₓ!: about 300,000 DOFs for Threads, 1,000–10,000 for Polyester. Threaded broadcast assignment (#357) uₕ .= a .* vₕ .+ wₕ into a VectorElement runs Base's own broadcast loop in bands under CpuThreaded and CpuPolyester, aliasing handled as Base does. Nested calls fall back to serial A Parallel() operation called from inside your own Threads.@threads loop, or from a task racing another one, runs its chunks serially instead of throwing Base's "@threads :static cannot be used concurrently or nested". Fixes Every Threads.@threads loop in the assembly sweeps uses the :static scheduler (#342). Documentation and benchmarks benchmark/policy_crossover.jl reports, on your own machine, where CpuThreaded and CpuPolyester start beating CpuSerial for each workload in 1D, 2D and 3D; the backend tutorial points to it. The CpuThreaded and CpuPolyester docstrings quote the measured crossovers; the internals page describes the threaded replay. Full changelog: https://github.com/gpena/Bramble.jl/compare/v3.12.0...v3.13.0

sigminer 2.3.3 Replaced future::availableCores() with parallelly::availableCores(). Fixed NULL row names in NMF matrix when using sig_tally() with Wang method, particularly for single-sample cases. F…

sigminer 2.3.3 Replaced future::availableCores() with parallelly::availableCores(). Fixed NULL row names in NMF matrix when using sig_tally() with Wang method, particularly for single-sample cases. Fixed compatibility issue with newer versions of SigProfilerExtractor by removing deprecated refit_denovo_signatures parameter from sigprofiler_extract(). The refit parameter is now deprecated and ignored. sig_estimate(), sig_extract() and bp_extract_signatures() can now use more than 2 cores: register a foreach backend to avoid the 2-core cap of the built-in parallel backend of the NMF package (#479). Fixed the "Unknown or uninitialised column" warnings of read_maf_minimal() when the input is a tibble (#461). sigminer 2.3.2

cgeniepy 0.16.0 handles cGENIE configurations beyond the modern 36x36x16 grid, and draws filled contours that reach the model's coastline. Other cGENIE configurations (#30) The grid helpers follow cG…

cgeniepy 0.16.0 handles cGENIE configurations beyond the modern 36x36x16 grid, and draws filled contours that reach the model's coastline. Other cGENIE configurations (#30) The grid helpers follow cGENIE's own formulas (GOLDSTEIN, BIOGEM, muffingen): get_genie_depth takes extra_levels (go_par_dk), and get_genie_lat takes equal_area=False (go_grid=1). Checked against runs with 17 levels, a 5500 m depth scale and longitude offsets of 0 and -180. to_geniebin(var, model=model) bins data onto a model's own grid, so the binned coordinates equal the model's. Longitude conversion works for any par_grid_lon_offset; normalise_longitude() now handles offset-0 grids. mask_basin works for any number of depth levels and time slices. Transect depth cells sit exactly on the model's zt_edges for any depth scale. Filled contours reach the coast (#33, fixes #31) With contourf=True, maps and transects are filled up to the model's coastline, the surface, the sea floor and the poles, with the land cells drawn on top. contourf=True turns pcolormesh off unless you pass pcolormesh=True, and contour lines follow the filled levels. Changes to existing results get_genie_depth(edge=False) now returns the zt that BIOGEM writes, which differs from the old arithmetic midpoints by up to 13 m. Depth labels from to_geniebin change with it. ArrComparison.crmse returns the centred RMSE; it used to return its square. Adding two GriddedData objects aligns them on their coordinates, where it used to add position by position. Fields stored in different longitude orders now sum correctly. Maps without a given ax use the Equal Earth projection instead of Eckert IV. contourf=True plots look different: see above. Fixes geniebin_depth binned nothing, and 3-D to_geniebin always failed. Passing N to geniebin_lat, geniebin_lon or normbin_lon failed. Plot cell edges come from the coordinates, so subsets such as lat=slice(-30, 30) are no longer stretched. Only the cell straddling a projection's edge is split (#4). GriddedData ** n no longer changes its operand. Plotting works with Matplotlib 3.11. mask_Arctic_Med is now documented as modern-36x36-only and warns for other shapes. Full changelog: https://github.com/ruiying-ocean/cgeniepy/compare/v0.15.2...v0.16.0

Code for "A structural causal framework for interventions on evolutionary accumulation models" (Diaz-Uriarte, Ríos-Arroyo, and Johnston; Royal Society Open Science, DOI 10.1098/rsos.261332; arXiv:2606…

Code for "A structural causal framework for interventions on evolutionary accumulation models" (Diaz-Uriarte, Ríos-Arroyo, and Johnston; Royal Society Open Science, DOI 10.1098/rsos.261332; arXiv:2606.12597).

In emerging 5G and beyond (B5G) network archi tectures, service-level agreement (SLA) assurance increasingly re quires systems that can detect deviations from expected behavior while providing interpr…

In emerging 5G and beyond (B5G) network archi tectures, service-level agreement (SLA) assurance increasingly re quires systems that can detect deviations from expected behavior while providing interpretable and uncertainty-aware evidence. Existing assurance frameworks commonly rely on deterministic reconstruction models to identify anomalous flows, treating recon struction error as a proxy for service degradation. However, such approaches provide limited information about the confidence of the detector and offer weak grounding for downstream explana tion and decision support. This paper introduces a framework for uncertainty-aware SLA violation detection in cloud-native 6G networks. The proposed approach utilizes a probabilistic au toencoder trained only on SLA-compliant traffic using Gaussian negative log-likelihood (NLL). The decoder outputs a feature wise reconstruction distribution parameterized by a mean and standard deviation, enabling each flow to be scored according to its likelihood under the learned compliant behavior. The resulting likelihood score is converted into an empirical anomaly percentile using the benign validation distribution, providing an interpretable measure of deviation severity. For each flagged flow, the framework also constructs structured evidence that can be consumed by cloud-native orchestration components to support downstream remediation and service assurance actions.

Preprint. Submitted to Computer Vision and Image Understanding on 5 September 2026; not peer reviewed. The PDF is the submitted manuscript, unchanged.Vision-language models fine-tuned for metric spati…

Preprint. Submitted to Computer Vision and Image Understanding on 5 September 2026; not peer reviewed. The PDF is the submitted manuscript, unchanged.Vision-language models fine-tuned for metric spatial reasoning on one family of sports fail to transfer to another: on basketball imagery, a racket-sport-tuned model underestimates distances as if the 28 m court were ~13 m and reports a ball in nearly every frame. These answer-distribution errors can be repaired without newly authored target QA or answer labels, using the data engine as a verifiable reward oracle: the engine that generates basketball questions from calibrated geometry can also score free-form answers. On an audited protocol with observable referents and instant-disjoint splits, one GRPO pass lifts distance T-MRA from 19.9 to 60.0 and reverses the counting regression. We audit that gain with baselines and controls RLVR studies omit: supervised fine-tuning on the same engine outputs reaches 61.7 at a prompt-matched budget and 71.5-72.4 at token- and update-matched budgets; a per-question constant fitted on the training split, with no image, scores 60.1; the RL policy keeps 40.3 with blanked and 53.5 with mismatched court images, while counting collapses without the image (59.0 vs. 26.0); and option-swapped relational pairs gain no pair-consistent accuracy from the image within noise. The pattern replicates on a second backbone (SmolVLM2-2.2B). Under the tested budgets, rewards, and GRPO implementation, engine-grounded adaptation repairs calibration: the answer distribution moves and metre-scale error shrinks, but 30 cm-level localization does not appear (acc@30cm <= 4.5; 0.0 under a precision-aligned reward), and RL shows no clear distance advantage over supervised consumption of that oracle. We release splits, scorers, predictions, and controls as a benchmark-and-audit package.Code, frozen protocols and evaluation artifacts: doi:10.5281/zenodo.22442707.

Ushbu maqolada yashil iqtisodiyot transformatsiyasi sharoitida barqaror iqtisodiy o‘sish sifatini baholashning nazariy-metodologik asoslarini takomillashtirish masalalari tadqiq etilgan. Tadqiqo…

Ushbu maqolada yashil iqtisodiyot transformatsiyasi sharoitida barqaror iqtisodiy o‘sish sifatini baholashning nazariy-metodologik asoslarini takomillashtirish masalalari tadqiq etilgan. Tadqiqotning asosiy maqsadi iqtisodiy o‘sishning miqdoriy ko‘rsatkichlari bilan bir qatorda uning ekologik barqarorligi, resurs va energiya samaradorligi hamda ijtimoiy natijadorligini kompleks baholashga yo‘naltirilgan metodologik yondashuvni shakllantirishdan iborat. Tadqiqotda tizimli, institutsional va qiyosiy tahlil, statistik guruhlash, indikatorlarni standartlashtirish hamda integral baholash usullaridan foydalanilgan. Natijada iqtisodiy, ekologik va ijtimoiy indikatorlarni o‘zaro integratsiyalash asosida barqaror iqtisodiy o‘sish sifatini baholashning kompleks yondashuvi ishlab chiqilgan. Tadqiqot natijalari yashil transformatsiya sharoitida iqtisodiy o‘sishning samaradorligini baholashda resurslardan oqilona foydalanish, uglerod sig‘imini kamaytirish, qayta tiklanuvchi energiya manbalarini kengaytirish, ekologik investitsiyalar va aholi farovonligi ko‘rsatkichlarini birgalikda hisobga olish zarurligini asoslaydi. Taklif etilgan metodologik yondashuv yashil iqtisodiyotga o‘tish jarayonlarining natijadorligini baholash va barqaror rivojlanish siyosatini ilmiy asoslashga xizmat qiladi.