Search Results (61)

We study the half-integer family $F(n,p)={}_3{F}_2\left(-\frac{2n+1}{2},1,\frac{2p-1}{2};\frac{2n+5}{2},\frac{2p+1}{2};1\right).$ Although this family belongs to the Karlsson–Minton class of ${}_3{F}_2\left(1\right)$ series with negative integral parameter differences, the most directly relevant symmetric formula of Shpot and Srivastava becomes singular at the free numerator parameter $a=1$(here a denotes the first numerator parameter of the Shpot–Srivastava reduction formula, the value of which controls the Beta factors in their evaluation). We show that this apparent singularity is removable and identify its finite value with the independently derived closed form. First, Euler’s integral representation and a recurrence for $J_{m,n}\left(x\right)={\int }_0^1{v}^{2m}{(1-x{v}^2)}^{-n}dv$ yield a closed form for ${}_2{F}_1\left(-\frac{2n+1}{2},1;\frac{2n+5}{2};x\right)$ as a polynomial in $(1-x)$ plus $arctanh\left(\sqrt{x}\right)/\sqrt{x}$. Rainville’s integral formula then reduces $F(n,p)$ to a rational part and a finite rational linear combination of odd harmonic sums $H_r^{\mathrm{odd}}={\sum }_{j=1}^r{(2j-1)}^{-1}$. For all integers $n\ge 0$ and $p\ge n+3$, this proves $F(n,p)\in \mathbb{Q}$. The threshold is sharp: at the adjacent boundary $p=n+2$ we obtain $F(n,n+2)=R_n+C_n{\pi }^2$ with $R_n,C_n\in \mathbb{Q}$ and $C_n\ne 0$, so rationality fails. We also prove, by analyticity and the independently derived formula, that the $a\to 1$ limit of the Shpot–Srivastava representation equals the present closed form. For computation, the exact assembly uses $O(n+p)$ arithmetic operations under the unit-cost model, while a direct fixed-precision evaluation is ill-conditioned for large n. A stable series/closed-form hybrid for the ${}_2{F}_1$ factor combined with adaptive positive-kernel Gauss–Jacobi quadrature gives relative errors at the machine-precision scale over the tested range up to $n=64$. Full article

Quantum federated learning (QFL) enables collaborative training of variational quantum circuits across decentralized clients, but client drift under non-IID data distributions degrades standard Federated Averaging (FedAvg). Existing aggregation methods either introduce tunable hyperparameters that depend on unknown data heterogeneity or incur structural costs such as order-dependent error propagation. We propose Kernel-Weighted Aggregation (KWA), a hyperparameter-free aggregation strategy that constructs a non-parametric kernel density estimate over the received client parameter vectors at each communication round. Clients in high-density regions receive greater voting power, and isolated clients are automatically attenuated. The kernel bandwidth is the median of all pairwise squared distances. We evaluate KWA against six state-of-the-art QFL methods and five classical robust baselines on five binary MNIST digit-pair tasks under IID, Dir($0.5$), and Dir($0.1$) distributions with a unified four-qubit benchmark. Under IID data, KWA recovers the performance of FedAvg. Under Dir($0.5$), KWA attains the highest average accuracy ($0.7120$), followed by FedAvg ($0.7093$). The margin at $N=4$ clients is not statistically significant. It grows to $+0.023$ at $N=16$ in the client scaling experiment. Under this distribution, the five classical robust baselines also rank below FedAvg. Under Dir($0.1$), KWA exceeds FedAvg by $0.028$ on average, with a pooled paired t-test $p=0.068$. Larger-circuit and three-class experiments show that the advantage does not yet transfer consistently beyond the four-qubit binary setting. KWA thus provides a hyperparameter-free aggregation strategy that recovers FedAvg under IID conditions, adapts to client drift, and requires no manually tuned hyperparameters. Full article

Formic acid (HCOOH), the most abundant carboxylic acid in the troposphere, modulates rainwater acidity, aerosol water uptake, and the oxidative capacity of remote atmospheres, yet its global budget remains poorly constrained. Herein, we present the first HCOOH total-column retrieval from thermal–infrared (TIR) measurements of the Thermal And Near-infrared Sensor for carbon Observation Fourier Transform Spectrometer-2 (TANSO-FTS-2) on board GOSAT-2, providing an early-afternoon observational perspective that complements existing morning low-Earth-orbit and geostationary HCOOH products. The Optimal Estimation retrieval sequentially fits the surface state, the atmospheric background (temperature, water vapor and ozone), and the HCOOH profile in a 1104–1109 cm⁻¹ microwindow centered on the ${\nu }_6$ Q-branch, with a radiance-ratio-scaled a priori that adapts to each scene. Averaging-kernel diagnostics concentrate the sensitivity in the 500–900 hPa layer with degrees of freedom for signal of approximately 1.05 under enhanced-emission conditions. For a 2019–2020 Australian bushfire case, including HCOOH in the state vector reduces the mean spectral residual from −0.327 K to 0.033 K. Independent evaluation against 113 time-coincident Toronto NDACC FTIR overpasses gives R = 0.95 and a zero-intercept slope of 2.12 for raw FTIR versus GOSAT-2. Applying the GOSAT-2 a priori and averaging kernel to the FTIR profiles changes the slope to 0.77 and reduces the RMSE to $0.23\times {10}^{16}$ molec cm⁻²; this one-sided smoothing is treated only as a sensitivity diagnostic. Monthly global maps for December 2019 and June 2020 show cross-sensor consistency with the IASI/MetOp-B ANNI-HCOOH product at R = 0.83 and 0.76. Over East Asia during April–June 2023, GOSAT-2 correlates with FY-4B/GIIRS at R = 0.90 (April) and R = 0.65 (June), with coherent three-sensor daily variability. These satellite comparisons are treated as cross-sensor consistency assessments rather than independent validation. GOSAT-2 consistently reports lower columns, a sensitivity-limited tendency consistent with a priori dominance under weak signals, limited information content, a narrow retrieval window, and differences among retrieval frameworks. The current product is a first demonstration for cloud-free daytime land scenes; this domain defines its sampling scope and representativeness but is not interpreted as a direct cause of the lower columns. The product offers a traceable GOSAT-2 TIR observational constraint on tropospheric HCOOH for future multi-platform synergy. Full article

Main memory carries data outside the processor’s trust boundary, so commodity systems-on-chip (SoCs) increasingly encrypt it; yet, in-line memory encryption engine itself becomes a differential power analysis (DPA) target whose key, if recovered, unlocks all of dynamic random-access memory (DRAM). We present PACE, a page-adaptive, cache-anchored memory encryption engine for RISC-V that makes nth-order DPA resistance practical and keeps cryptographic latency off the cache eviction critical path. PACE inserts a TileLink adapter between the last-level cache and the memory port and applies, per physical page, one of four policies (plaintext/confidentiality/confidentiality+integrity/+masking-order-d) selected from RISC-V page table bits through a memory-mapped control plane. Confidentiality uses counter mode whose per-line keystream is precomputed during cache residency; integrity is tree-free at the embedded operating point via on-chip counters and tags, with a live split counter block-MAC Bonsai Merkle tree for scale-out. DPA resistance is layered: ISAP-style fresh re-keying caps the data complexity per key at $q\approx 1$, and domain-oriented masking (DOM, d + 1 shares) protects the sole key processing block to order d. We implement PACE in Chisel on a Rocket SoC (Chipyard) and evaluate it with open-source tooling. A deterministic TileLink-level harness proves ciphertext-in-memory and detects tamper/replay/splice, and the live Tier-B engine (DRAM counters and per-line message authentication codes (MACs) plus an on-chip-rooted block-MAC tree) is validated from end to end on full Rocket and BOOM SoCs and on the FPGA; the masked Ascon-p S-box is proven order-d secure (d = 1, 2) under a glitch- and transition-aware model by three independent formal tools (COCO, PROLEAD, and SILVER, the last also deciding the full composability lattice and confirming exact glitch-robust order-2 probing security), with COCO extending the exact verdict to the highest synthesized order d = 3 (secure at probing orders 1–3); a simulated trace correlation power analysis (CPA) recovers the full key from an unprotected core and is defeated by masking, with a mutual information analysis confirming the $N\propto {\sigma }^{2(d+1)}$ trace amplification law. We further realize PACE on field-programmable gate array (FPGA) silicon: the engine plus an on-chip ring oscillator power sensor is placed, routed, timing-closed at 100 MHz, and programmed on a Xilinx XC7Z020, and we drive a fixed-vs-random Test Vector Leakage Assessment (TVLA) campaign read back entirely over a JTAG (Joint Test Action Group). A multi-core configuration and a Linux control-plane driver are likewise validated. Across synthetic access patterns and named application kernels (AES, SHA-256, matrix multiplication, pointer chasing) on both in-order Rocket and out-of-order BOOM, application-level overhead is within measurement noise of plaintext for cache resident workloads (masking, in particular, is cycle-identical to plain confidentiality), and we characterize the cost of each policy, masking order, and re-keying interval, demonstrating side-channel-hardened memory encryption on open RISC-V hardware. Full article

Aiming at the problems of low facial expression recognition accuracy, high computational complexity, and difficulty in edge deployment caused by variable illumination, diverse head poses, severe partial occlusions, and the long-tail distribution of expression categories in complex power business hall scenarios, this paper presents an engineering-oriented lightweight dynamic perception facial expression recognition method for complex power business hall scenarios. The proposed method adopts an end-to-end joint face detection and expression classification framework built upon the anchor-free CenterNet architecture. In terms of technical implementation, this paper primarily focuses on the integration and adaptation of existing techniques to meet scenario-specific requirements; integrates a specially designed Multi-Scale Fusion Deformable Large Kernel Attention (MSF-DLKA) module to enhance multi-scale perception of subtle expression deformations under varying poses; designs a Task-Aware Dynamic Detection Head (TADDH) with decoupled spatial-channel attention to separately adapt the feature requirements of localization and classification subtasks; directly employs the off-the-shelf Label-Distribution-Aware Margin loss (LDAM Loss) to alleviate class imbalance; and combines three established compression techniques—structured pruning, quantization-aware training, and knowledge distillation—into a progressive lightweight pipeline for edge deployment. Experiments on two public micro-expression datasets, CASME II and SAMM, as well as a self-built electric power business hall scenario dataset (PBHD), show that the proposed method achieves mAP@0.5 of 93.2%, 93.9%, and 91.7%, and macro-F1 of 0.929, 0.936, and 0.914 on the three datasets, respectively, while using only 13.5 M parameters and attaining a single-image inference time of 28.9 ms on the NVIDIA Jetson AGX Orin. The paper provides a feasible solution for real-time expression perception in resource-constrained scenarios such as power business halls. Full article

Protective reinforced-concrete (RC) elements designed to resist contact blast loading must reconcile high energy dissipation with material and weight efficiency. This study examines HollowHex, an RC slab architecture in which periodic hexagonal cellular voids redistribute blast-induced stresses along inclined web-walls through a Vierendeel-type framing action. A full-factorial design of experiments across web thickness, charge mass, and hexagonal cell radius was carried out with Abaqus/Explicit using a concrete-damaged-plasticity model and mass-dependent Friedlander overpressure histories calibrated to UFC 3-340-02 scaled-distance relations. A six-level mesh-convergence study with three independent fine-mesh verification runs established the residual mesh effect as regime-dependent, bounded within approximately 13% in the elastic and severe-damage regimes and approximately 18% in the transition regime. Ten surrogate-model families—linear, polynomial, kernel, ensemble, and multilayer-perceptron—were benchmarked under leave-one-out, 5-fold, and 7-fold cross-validation. The best models achieved out-of-sample R² = 0.96 for peak displacement and R² = 0.93 for a continuous damage volume ratio (DVR), with train-to-validation gaps of only 0.03 and 0.06, indicating genuine generalization on the small dataset. A direct identical-condition comparison against circular-hollow slabs of matched void area shows blast-equivalent performance across the elastic, transition, and severe damage regimes (peak displacements within 2%, damage volume ratios within 7%), positioning the hexagonal architecture as a blast penalty-free alternative whose selection can be driven by non-blast criteria. A cross-validated parametric design heatmap is provided as a screening tool within the verified envelope. The uniform loading idealization is cross-checked against the spatially resolved CONWEP model, conservative on peak displacement by a factor of approximately 3.5, while approximately damage-equivalent and the constitutive model is validated at the damage level against documented contact-explosion tests through coupled FEM–SPH simulation. The findings position HollowHex not as a universally superior geometry but as a quantitatively beneficial alternative within the service/transition design range of greatest practical interest for blast protection. Full article

This work constructs a family of weighted higher-order Opial-type estimates on an arbitrary time scale $\mathbb{T}$, with the weight structure generated by a non-decreasing auxiliary function $\omega$ and its delta derivative ${\omega }^{\Delta }$. For n-times delta differentiable functions whose delta derivatives of orders $0,1,\dots ,n-1$ vanish at the left endpoint, the mixed functional ${\int }_a^b{v\left(t\right)|u\left(t\right)|}^p{|u^{{\Delta }^n}\left(t\right)|}^q\Delta t$ is bounded by expressions involving only the highest-order delta derivative $u^{{\Delta }^n}$. Three forms are obtained: a Hölder-type product estimate, a Young-type two-term estimate with a free balancing parameter $\theta >0$, and an optimized single-integral estimate with the explicit constant $K={(r-1)}^{1-1/r}/r$, where $r=(p+q)/q$. The proofs rely on the Taylor representation on time scales, Hölder’s inequality, Jensen’s inequality, and Fubini’s theorem. The resulting bounds are governed by computable kernels that encode the interaction between the Taylor monomials, the auxiliary weight, and the external weight. Specializations to $\mathbb{T}=\mathbb{R}$, $\mathbb{T}=\mathbb{Z}$, and the quantum lattice $q_0^{{\mathbb{N}}_0}$ recover classical, discrete, weighted, and quantum Opial inequalities with their standard constants and also yield several higher-order weighted estimates. As an application, a uniqueness criterion for higher-order dynamic initial value problems is established. Full article

Rolling-element bearings remain a leading cause of unplanned downtime in industrial machinery, while vibration-based condition monitoring has matured, the post-2018 literature has converged on machine-learning classifiers whose interpretability layer remains restricted to expert analysts. This paper presents an operator-centred visualization framework supported by two time-frequency methods: the Zhao–Atlas–Marks Distribution (ZAMD), a Cohen’s-class representation with a cross-term-suppressing cone kernel, and Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN), evaluated through its Hilbert spectral analysis output. Both methods produce two-dimensional time-frequency artefacts with a similar visual structure—impact-related energy bursts that recur at the characteristic fault frequencies—and are presented in side-by-side form for each fault class. A four-stage framework wraps either method with the characteristic fault frequencies (supplied as a comparison reference) and colour-coded, healthy baseline-referenced scaling. The framework is demonstrated on a laboratory bearing rig (KOYO 6302, 600 RPM) across inner-race, outer-race, and ball-spin fault classes. A preliminary readability assessment of annotated ZAMD-generated artefacts, with twelve non-specialist participants from a brewing and packaging industrial context, recorded 89.8% aggregate classification accuracy (194 of 216 trials) at a mean response time of 15.4 s. Because no label-free or alternative-format control conditions were included, this result characterises the annotated artefact as a whole and does not isolate the contribution of the time-frequency representation from that of the annotation layer; it is established for the ZAMD engine only. The two methods are compared as visualization engines—qualitatively, through the structure of their side-by-side time-frequency artefacts, and quantitatively, through computational cost—whereas the non-specialist readability assessment characterises the ZAMD-based framework specifically. CEEMDAN is positioned as a candidate alternative engine whose time-frequency output is shown to be structurally similar but whose operator readability has not been tested with human participants and is identified as future work. Full article

Visibility graphs map time series to networks whose combinatorial structure encodes fractality, recovering the Hurst exponent of self-affine processes. We ask what the visibility construction carries beyond this fractal content. We lift the visibility graph to a 1-varifold, a measure on position and direction space from geometric measure theory, and equip it with a multiscale positive definite kernel. The lift embeds visibility graphs of unequal size in a common Hilbert space and yields a channel-resolved measure of cross-series geometric alignment. On a 25.8-year daily panel of thirteen commodity and currency layers, we define a relative alignment contrast that compares commodity currencies and safe haven currencies in their geometric alignment with the commodity complex. During global risk-off episodes the contrast is large and positive: commodity currencies import commodity shock geometry far beyond a time-shift independence benchmark, while the Japanese yen remains near geometric independence and the franc is confounded by a managed regime. The contrast is significant under three stress definitions with autocorrelation robust inference, holds as a continuous dose response, survives the removal of any single crisis, withstands moment, fractal, and topological controls, is direction-consistent across sixteen specifications, and collapses under a time-shift placebo. Detrended fluctuation analysis explains only two percent of it, so the reconfiguration is geometric information beyond fractality at this horizon, and a scaling exponent of the kernel mass separates a fractal-free component from a fractal-driven one. For investors, financial institutions, and policymakers, the contrast is a real-time structural diagnostic of flight to safety: it marks when commodity currencies stop diversifying the commodity complex while genuine safe havens still do, signaling through a channel that second-moment risk measures are built to miss. Full article

Photoacoustic spectroscopy (PAS) is a highly sensitive and non-destructive technique widely used for trace gas detection; however, the simultaneous quantification of methane (CH₄), ethane (C₂H₆), and ethylene (C₂H₄) remains challenging due to severe spectral cross-interference and non-linear responses across broad concentration ranges. In this work, we propose a high-precision, end-to-end detection framework based on a Deep Kernel Extreme Learning Machine (DKELM) optimized using a Mutation–Chaotic Particle Swarm Optimization (MCPSO) algorithm. To enhance diagnostic information in the photoacoustic signals, a multi-scale wavelet transform based on a db4 wavelet basis with 5-layer decomposition and a Heursure soft threshold strategy is first employed for denoising and enhancing absorption features. To address the hyperparameter sensitivity and local-optimum trapping inherent in deep models, the MCPSO algorithm integrates hybrid chaotic initialization, adaptive mutation probability control, Cauchy-based perturbation, temperature-controlled mutation amplitude, and elite-guided population updating. The proposed MCPSO-DKELM model is evaluated on an expanded dataset of 470 mixed-gas spectra and benchmarked against other frameworks, including the previously reported SVM-CPSO-KELM architecture. The experimental results demonstrate that MCPSO-DKELM achieves stable, segmentation-free quantification across the full dynamic range, with an average detection error below 3.5% and the maximum relative error constrained to under 15%, which represents a substantial improvement over existing approaches. Thus, the combination of deep kernel feature extraction and mutation–chaotic global optimization provides a robust and reliable solution for simultaneous multi-component hydrocarbon gas analysis in complex industrial environments. Full article

Deterministic radio wave propagation models provide high accuracy in complex outdoor environments but remain computationally impractical for large-scale network planning and spectrum management. In contrast, empirical and hybrid models offer low complexity at the expense of reduced accuracy, systematic bias, and limited terrain sensitivity. This paper proposes a unified delta learning framework that enhances fast baseline propagation models by learning a data-driven correction toward a deterministic Parabolic Equation Modeling (PEM) reference. A key novelty lies in a compact, physics-informed feature representation that replaces the full terrain profile with an 18-dimensional vector combining local geometric descriptors, global terrain characteristics, and baseline responses, enabling accurate correction with low-dimensional input. The study also provides the first systematic investigation of delta-based correction across multiple widely used propagation models. The framework is evaluated for free-space propagation, ITU-R P.1546, ITU-R P.1812, and ITU-R P.452 using ridge regression, kernel ridge regression, gradient boosting regression trees, and a neural network model. Model performance is assessed in terms of error reduction, bias mitigation, robustness across learning algorithms, and profile-level generalization to previously unseen propagation paths within the considered terrain categories. Results show substantial error reduction, with up to twofold improvement for simpler baseline models and consistent gains for hybrid models, while preserving computational efficiency. Full article

Hereditary damping and fractional attenuation are widely used to model wave propagation in complex media, but the variational and spectral origin of the corresponding nonlocal-in-time operators is often left implicit. In this work, we derive such operators from a minimal conservative field–reservoir model. A real scalar field is coupled locally to a continuum of harmonic reservoir modes, which are then eliminated exactly. The resulting reduced dynamics is a causal wave equation with a memory-friction term acting on the field velocity. The memory kernel is generated by the reservoir coupling spectrum through a cosine-transform relation, establishing a direct spectrum-to-kernel correspondence. This relation provides both a physical interpretation of hereditary damping and a practical admissibility criterion: macroscopic attenuation and dispersion arise from the delayed back-action of unresolved internal modes, while physically admissible kernels are constrained by the non-negativity of the underlying spectral density. The framework unifies several standard damping regimes. A broadband reservoir recovers the Markovian locally damped wave equation, reservoirs with a finite characteristic time generate finite-memory relaxation and frequency-dependent dispersion, and scale-free reservoir spectra produce power-law memory kernels. In the latter case, the hereditary damping operator reduces to a Caputo-type fractional derivative, showing that fractional wave attenuation can emerge as an effective reduced dynamics rather than being postulated phenomenologically. We further analyze dispersion, attenuation, causality, stability, and admissibility conditions in terms of the reservoir spectrum. The main contribution of the work is therefore to provide a variational and spectral derivation of hereditary and fractional wave damping, linking the structure of unresolved reservoir modes to macroscopic nonlocal wave dynamics. Full article

Convolutional neural networks (CNNs) and training-free CNN variants have been successfully applied to hyperspectral image (HSI) processing and analysis. Training-free CNNs have shown promising feature extraction performance, which could effectively address the issue of typical CNNs being highly parameterized; however, inevitable noise and redundancy in the randomly selected training-free convolutional kernels often leads to unsatisfactory performance. To address this issue, we propose Superpixel Random Selection Random Walk Multi-Branch Depthwise Convolutional Neural Network (SRSRWMD-CNN). Specifically, we propose a novel training-free convolutional neural network characterized by inter-layer multi-scale integration and intra-layer grouping. Various superpixels groups are first generated through multi-scale superpixel segmentation algorithms, then the predetermined number of superpixels are randomly sampled from these groups to serve as training-free convolution kernels. This mechanism enables adaptive computation of HSI feature maps without costly model training in the feature extraction stage, allowing the network to effectively capture a multi-scale spectral–spatial feature representation. Additionally, we propose a multi-branch depthwise convolution strategy that mitigates feature learning errors while significantly enhancing feature representation capabilities. A random walk strategy is employed to expand the receptive field and enhance the robustness of the training-free convolution kernels. Finally, the multi-scale spectral–spatial features are concatenated with the multiple convolutional stages to fuse salient shallow and deep features for accurate HSI classification. Extensive experiments demonstrate that the proposed method achieves superior performance compared to state-of-the-art algorithms. Full article

Existing CNN-based visual malware classification methods are often constrained by inductive bias mismatch: standard isotropic convolution kernels and global pooling operations neglect the inherent structural anisotropy of malware images, and these methods struggle to address the spatial rearrangement of code blocks caused by obfuscation, which we term the “Malware Picasso Problem”. To overcome these limitations, we propose AD-CapsFPN, an end-to-end framework representing a significant step toward spatial reasoning over texture memorization, with a synergistic “Rectification–Fusion–Inference” mechanism. Our approach rectifies anisotropic inductive biases in the feature extraction stage, dynamically aggregates cross-scale discriminative features in intermediate layers, injects row-aware spatial biases, and adopts a global pooling-free spatial routing strategy in the classification stage, effectively reconstructing logical associations between obfuscated and scattered code blocks. Experiments on the large-scale Fusion dataset and the obfuscated Androdex dataset demonstrate significant performance improvements: our method achieves a 16.22% boost in macro F1-score over the MobileNetV4 baseline on the Fusion dataset (reaching 97.98%), and hits 92.45% macro F1-score on the highly challenging Androdex-Set1, outperforming state-of-the-art methods such as MDC-RepNet (88.97%) and TAEfficientNet (88.15%). This work confirms that embedding malware domain priors into architecture design is the key to robust malware classification. Full article

We present a unified GPU-accelerated framework for real-time Smoothed Particle Hydrodynamics (SPH) fluid simulation with two-way rigid body coupling, secondary particle effects, and volumetric rendering, implemented entirely within the Unity game engine. The framework employs a weakly compressible SPH formulation with O(n) count sort-based spatial hashing and introduces a signed distance field (SDF) coupling system that evaluates three representative geometric primitives, sphere, cylinder, and torus, of increasing topological complexity directly on the GPU. Bidirectional force exchange is achieved through lock-free atomic compare-and-swap impulse accumulation, enabling thousands of fluid particles to interact simultaneously with each rigid body without serialization. A GPU stream compaction–based secondary particle system generates and classifies foam, spray, and bubble effects in real time, while a volumetric rendering pipeline samples fluid density into a 3D texture for SDF-composited volume rendering without surface mesh extraction. A conditional kernel dispatch strategy eliminates GPU cycles for disabled subsystems, and dynamic buffer management reduces memory pressure through runtime allocation. The system sustains above 54 frames per second at four million particles on a consumer-grade GPU, with sub-linear frame time scaling and a 1.70× speedup from dynamic buffer allocation over static pre-allocation. Full article