Compression Method Research Articles

Investigating the Influence of Crospovidone's Manufacturer Variability on Dissolution Profiles of Hydrochlorothiazide Tablets.

This study examines the influence of crospovidone (CP) manufacturer variability on the dissolution profiles of hydrochlorothiazide (HCTZ) tablets. Four CP batches from different manufacturers were characterized using pharmacopeial and physicochemical tests, including infrared absorption, loss on drying, and scanning electron microscopy (SEM). Significant differences were observed in the particle size distribution, wetting time, and water absorption capacities of the CP batches. Tablets were formulated using both direct compression and wet granulation methods. For the latter, the superdisintegrant was either added to the binder solution or incorporated intra- or extra-granularly. Disintegration and dissolution tests revealed that both CP concentration and the method of incorporation significantly affected tablet performance. Poly Kovidone and Max-Povidon exhibited superior performance at lower concentrations, while differences between brands became less pronounced at higher concentrations. The extra-granular method notably enhanced drug release profiles. Statistical analyses, including f2 similarity factors and MANOVA with Principal Component Analysis (PCA), highlighted significant differences in dissolution behavior among the formulations. These findings emphasize the importance of controlling excipient variability to ensure consistent product performance. The study concludes that a 2% CP concentration is optimal for mitigating source variability and that the extra-granular addition of CP in wet granulation is recommended for enhancing its functional properties.

Full reference point cloud quality assessment using support vector regression

Point clouds are a general format for representing realistic 3D objects in diverse 3D applications. Since point clouds have large data sizes, developing efficient point cloud compression methods is crucial. However, excessive compression leads to various distortions, which deteriorates the point cloud quality perceived by end users. Thus, establishing reliable point cloud quality assessment (PCQA) methods is essential as a benchmark to develop efficient compression methods. This paper presents an accurate full-reference point cloud quality assessment (FR-PCQA) method called full-reference quality assessment using support vector regression (FRSVR) for various types of degradations such as compression distortion, Gaussian noise, and down-sampling. The proposed method demonstrates accurate PCQA by integrating five FR-based metrics covering various types of errors (e.g., considering geometric distortion, color distortion, and point count) using support vector regression (SVR). Moreover, the proposed method achieves a superior trade-off between accuracy and calculation speed because it includes only the calculation of these five simple metrics and SVR, which can perform fast prediction. Experimental results with three types of open datasets show that the proposed method is more accurate than conventional FR-PCQA methods. In addition, the proposed method is faster than state-of-the-art methods that utilize complicated features such as curvature and multi-scale features. Thus, the proposed method provides excellent performance in terms of the accuracy of PCQA and processing speed. Our method is available from https://github.com/STAC-USC/FRSVR-PCQA.

Neural-WDRC: A Deep Learning Wide Dynamic Range Compression Method Combined With Controllable Noise Reduction for Hearing Aids.

Wide dynamic range compression (WDRC) and noise reduction both play important roles in hearing aids. WDRC provides level-dependent amplification so that the level of sound produced by the hearing aid falls between the hearing threshold and the highest comfortable level of the listener, while noise reduction reduces ambient noise with the goal of improving intelligibility and listening comfort and reducing effort. In most current hearing aids, noise reduction and WDRC are implemented sequentially, but this may lead to distortion of the amplitude modulation patterns of both the speech and the noise. This paper describes a deep learning method, called Neural-WDRC, for implementing both noise reduction and WDRC, employing a two-stage low-complexity network. The network initially estimates the noise alone and the speech alone. Fast-acting compression is applied to the estimated speech and slow-acting compression to the estimated noise, but with a controllable residual noise level to help the user to perceive natural environmental sounds. Neural-WDRC is frame-based, and the output of the current frame is determined only by the current and preceding frames. Neural-WDRC was compared with conventional slow- and fast-acting compression and with signal-to-noise ratio (SNR)-aware compression using objective measures and listening tests based on normal-hearing participants listening to signals processed to simulate the effects of hearing loss and hearing-impaired participants. The objective measures demonstrated that Neural-WDRC effectively reduced negative interactions of speech and noise in highly non-stationary noise scenarios. The listening tests showed that Neural-WDRC was preferred over the other compression methods for speech in non-stationary noises.

Several compressible computational fluid dynamics methods applied to transient sheet/cloud cavitation

This paper introduces several compressible computational fluid dynamics (CFD) methods and assesses their ability to simulate typical sheet-to-cloud cavitating flow around a hydrofoil. More precisely, the Tait equation of state is used to describe the density of water, while the ideal gas equation of state is used to model the density of vapor. The first method assumes that the cavitation is a multiphase flow with isothermal conditions, meaning that it exhibits isothermal compressibility. Based on the first method, the second and third methods take into account the thermal energy and total energy equations, respectively, i.e., the thermal energy compressibility and the total energy compressibility. An incompressible simulation is also performed for the comparison. The results show that all of the strategies successfully replicate the periodic breakup of the sheet cavity and the formation of the cloud cavity. The predicted frequency of cavity shedding using compressible methods is higher than that using the incompressible method. In addition, all the CFD simulations confirm that the disturbance moving upward in the sheet cavity is actually a condensation shock. The overpressure resulting from the collapse of the cavity can be captured using three compressible approaches. The boundary layer and time-averaged hydrofoil pressure coefficient are compared and analyzed, revealing a negligible difference among the three compressible simulation results.

A Novel Context-Aware Douglas–Peucker (CADP) Trajectory Compression Method

Most traditional trajectory compression methods, such as the Douglas–Peucker (DP) method, consider only spatial characteristics and disregard contextual factors, including environmental context. This paper proposes a new way of trajectory formulation by considering all spatial, internal, environmental, and semantic contexts to capture all contextual aspects of moving objects. Then, we propose the Context-Aware Douglas–Peucker (CADP) method for trajectory compression. These facts are confirmed by experiments with real AIS data showing that, while CADP preserves the same computational efficiency of DP (i.e., at O(n2)), it outperforms DP and two-stage Context-Aware Piecewise Linear Segmentation (two-stage CPLS) methods in preserving agent movement behavior, obtaining compressed trajectories that are closer to the original ones and that are much more useful in base analyses such as trajectory prediction. Specifically, the LSTM-based models trained on CADP-compressed trajectories have relatively lower RMSEs than others compressed by either DP or two-stage CPLS. Therefore, CADP is more scalable and efficient, thus making it more practical for large-scale engineering applications; with the improvement in trajectory analysis accuracy achieved by the suggested method, a wide range of critical engineering applications can be potentially improved, such as collision avoidance and route planning. Future work will focus on spatial auto-correlation and uncertainty to extend the robustness and applicability of the approach.

Application of Karhunen-Loève Transform Algorithms in Hydrogeological Parameter Compression: A Comparative Study

ABSTRACT This paper proposes a parameter compression method using the Karhunen-Loève Transform (K-L) to reduce the computational burden of parameter inversion in groundwater models. Two spatial distributions of hydraulic conductivity (K) were generated using Kriging interpolation and Sequential Gaussian Simulation (SGS), and the K-L transform was applied to reduce their dimensionality. The method’s generality was assessed by varying model boundaries, extreme values of K, and simulation periods. The results showed that sequentially adding different boundary conditions resulted in high accuracy in model output. The model was most accurate when K values were within the range of 50-100 m/d, and its performance decreased as the range of K expanded. The model also exhibited high accuracy across various time periods within a single stress period. Dimensionality reduction through Kriging-KL and SGS- KL resulted in high accuracy of model-simulated heads, with R2 values exceeding 0.80 and 0.98 respectively. Overall, the K-L transform efficiently reduces parameter dimensions without compromising model performance. SGS-KL showed greater robustness under varying parameters, indicating its suitability for complex groundwater systems. This study contributes to efficient groundwater model inversion and provides valuable insights into the characteristics of groundwater systems.

Computation of Viscoelastic Shear Shock Waves Using Finite Volume Schemes With Artificial Compressibility.

The formation of shear shock waves in the brain has been proposed as one of the plausible explanations for deep intracranial injuries. In fact, such singular solutions emerge naturally in soft viscoelastic tissues under dynamic loading conditions. To improve our understanding of the mechanical processes at hand, the development of dedicated computational models is needed. The present study concerns three-dimensional numerical models of incompressible viscoelastic solids whose motion is analysed by means of shock-capturing finite volume methods. More specifically, we focus on the use of the artificial compressibility method, a technique that has been frequently employed in computational fluid dynamics. The material behaviour is deduced from the Fung-Simo quasi-linear viscoelasiticity (QLV) theory where the elastic response is of Yeoh type. We analyse the accuracy of the method and demonstrate its applicability for the study of nonlinear wave propagation in soft solids. The numerical results cover accuracy tests, shock formation and wave focusing.

FORMULATION AND OPTIMIZATION OF GASTRO-RETENTIVE FLOATING TABLETS OF ENALAPRIL MALEATE AND LOSARTAN FOR ENHANCED BIOAVAILABILITY AND THERAPEUTIC EFFICACY

The objective of this study was to formulate and optimize gastro-retentive floating tablets of Enalapril Maleate and Losartan to enhance their bioavailability and therapeutic efficacy. These drugs, commonly used in the management of hypertension and heart failure, suffer from limited solubility and rapid gastric emptying, which reduce their effectiveness. Gastro-retentive floating tablets were developed using hydroxypropyl methylcellulose (HPMC K4M) and guar gum as release-controlling agents, with sodium bicarbonate and citric acid as effervescent agents to ensure buoyancy. Twenty formulations were prepared using the direct compression method and evaluated for critical parameters, including hardness, friability, swelling index, floating lag time, total floating time, and in-vitro drug release.Preformulation studies confirmed the compatibility of the drugs with excipients, with no chemical interactions detected using FTIR spectroscopy. The micrometric properties of the powder blends demonstrated excellent flowability and compressibility, ensuring suitability for tablet manufacturing. All formulations showed acceptable weight variation, hardness, and friability. The floating lag time ranged from 4.3 to 5.1 minutes, and total floating times exceeded 8 hours for all batches. Drug content ranged from 96.79% to 99.57%, reflecting uniformity and accuracy in dosing.In-vitro dissolution studies revealed sustained release profiles, with optimized formulations achieving over 90% cumulative drug release within 12 hours. Batch F7 demonstrated the best overall performance, with a high swelling index, minimal floating lag time, prolonged floating duration, and consistent drug release. These findings indicate that gastro-retentive floating tablets of Enalapril Maleate and Losartan can effectively address challenges associated with poor bioavailability and rapid gastric emptying, offering a promising approach for improved therapeutic outcomes and patient compliance.

Scaling down for efficiency: Medium-sized protein language models perform well at transfer learning on realistic datasets.

Protein language models (pLMs) can offer deep insights into evolutionary and structural properties of proteins. While larger models, such as the 15 billion parameter model ESM-2, promise to capture more complex patterns in sequence space, they also present practical challenges due to their high dimensionality and high computational cost. We systematically evaluated the performance of various pLMs across multiple biological datasets to assess the impact of model size on transfer learning. Surprisingly, we found that larger models not necessarily outperform smaller ones, in particular when data is limited. Medium-sized models, such as ESM-2 650M and ESM C 600M, demonstrated consistently good performance, falling only slightly behind their larger counterparts-ESM-2 15B and ESM C 6B-despite being many times smaller. Additionally, we compared various methods of compressing embeddings prior to transfer learning, and we found that mean embeddings consistently outperformed other compression methods. In summary, ESM C 600M with mean embeddings offers an optimal balance between performance and efficiency, making it a practical and scalable choice for transfer learning in realistic biological applications.

Preparation of Hydroxyapatite-Aligned Collagen Sheets and Their Evaluation for Fibroblast Adhesion and Collagen Secretion.

The structure of many native tissues consists of aligned collagen (Col) fibrils, some of which are further composited with dispersed hydroxyapatite (HAp) nanocrystals. Accurately mimicking this inherent structure is a promising approach to enhance scaffold biocompatibility in tissue engineering. In this study, biomimetic sheets composed of highly aligned Col fibrils were fabricated using a plastic compression and tension method, followed by the deposition of HAp nanocrystals on the surface via an alternate soaking method. The fabricated Col sheets exhibited high solid density, retained the native periodicity (D-band) of Col fibrils, and displayed plate-like HAp nanocrystals dispersed on their surface. In vitro experiments demonstrated that these sheets could regulate fibroblasts adhesion, inducing more elongated nuclei and oriented actin bundles on the aligned Col sheets. Analysis of focal adhesion assembly revealed greater cell focal adhesions on the aligned composite sheets compared to those with random Col fibril structures. Fibroblasts cultured on aligned Col with partly HAp-mineralized sheets exhibited the highest cell-extracellular matrix (ECM) protein secretion, highlighting that HAp incorporation and fibroblast alignment synergistically promote early ECM formation and wound healing. These results suggest that highly aligned Col fibrils with dispersed HAp nanocrystals, closely mimicking the microarchitecture of natural tissues, have significant potential to control cell adhesion and protein secretion for tissue engineering applications.

Direct compression tablet formulation of trimetazidine through systematic screening of oxalate salts.

Salt formation has been extensively used to modulate and improve the properties of active pharmaceutical ingredients (API), such as solubility, stability and mechanical properties. Tablets of the anti-angina drug, trimetazidine (TMZ) are currently manufactured using the wet granulation process, rather than the more cost-effective direct compression method. In an effort to address the two main challenges associated with the commercial dihydrochloride salt (TMZ-2HCl), i.e., poor flowability and high hygroscopicity, we have screened and prepared three new oxalate (Oxa) salts of TMZ, i.e., TMZ:Oxa 1:1 (mono-salt), 2:1 (hemi-salt) and 1:2 (di-salt). All of the three salts exhibited improved solid-state properties over TMZ-2HCl, including flow, tabletability, and stability against high humidity. Among the three salts, the TMZ-Oxa (1:2) di-salt demonstrates overall superior properties, establishing it as the most promising alternative crystal form. Consequently, using the di-salt, we developed a direct compression tablet formulation that shows potential for commercial manufacturing.

CAFE+: Towards Compact, Adaptive, and Fast Embedding for Large-scale Online Recommendation Models

The growing memory demands of embedding tables in Deep Learning Recommendation Models (DLRMs) pose great challenges for model training and deployment. Existing embedding compression solutions cannot simultaneously achieve memory efficiency, low latency, and adaptability to dynamic data distribution. This paper presents CAFE+, a Compact, Adaptive, and Fast Embedding compression framework that meets the above requirements. The design philosophy of CAFE+ is to dynamically allocate more memory to important features, and less to unimportant ones. We assign unique embedding to important feature and allow multiple unimportant features sharing one embedding. We propose a fast and lightweight feature monitor, to real-time capture feature importance and report important features. We theoretically analyze the accuracy of our feature monitor, and prove the superiority of CAFE+ from the aspect of model convergence. Extensive experiments show CAFE+ outperforms existing embedding compression methods, yielding \(3.94\%\) and \(3.94\%\) superior testing AUC on Criteo Kaggle dataset and CriteoTB dataset at a compression ratio of \(10000\times\) . Building on our conference version [114], this journal version introduces several novel designs (Implicit Importance Attenuation, Adaptive Threshold Adjustment, and ColdSifter) that enable CAFE+ to more effectively adapt to long-term online learning and achieve better model quality. All codes are available at GitHub [112].

A lossless reference-free sequence compression algorithm leveraging grammatical, statistical, and substitution rules.

Deoxyribonucleic acid (DNA) or ribonucleic acid (RNA) sequence compressors for novel species frequently face challenges when processing wide-scale raw, FASTA, or multi-FASTA structured data. For years, molecular sequence databases have favored the widely used general-purpose Gzip and Zstd compressors. The absence of sequence-specific characteristics in these encoders results in subpar performance, and their use depends on time-consuming parameter adjustments. To address these limitations, in this article, we propose a reference-free, lossless sequence compressor called GraSS (Grammatical, Statistical, and Substitution Rule-Based). GraSS compresses sequences more effectively by taking advantage of certain characteristics seen in DNA and RNA sequences. It supports various formats, including raw, FASTA, and multi-FASTA, commonly found in GenBank DNA and RNA files. We evaluate GraSS's performance using ten benchmark DNA sequences with reduced number of repeats, two highly repetitive RNA sequences, and fifteen raw DNA sequences. Test results indicate that the weighted average compression ratios (WACR) for DNA and RNA sequences are 4.5 and 19.6, respectively. Additionally, the entire DNA sequence corpus has a total compression time (TCT) of 246.8 seconds (s). These results demonstrate that the proposed compression method performs better than several advanced algorithms specifically designed to handle various levels of sequence redundancy. The decompression times, memory usage, and CPU usage are also very competitive. Contact: anirban@klyuniv.ac.in.

Lossless and reference-free compression of FASTQ/A files using GeneSqueeze

As sequencing becomes more accessible, there is an acute need for novel compression methods to efficiently store sequencing files. Omics analytics can leverage sequencing technologies to enhance biomedical research and individualize patient care, but sequencing files demand immense storage capabilities, particularly when sequencing is utilized for longitudinal studies. Addressing the storage challenges posed by these technologies is crucial for omics analytics to achieve their full potential. We present a novel lossless, reference-free compression algorithm, GeneSqueeze, that leverages the patterns inherent in the underlying components of FASTQ files to solve this need. GeneSqueeze’s benefits include an auto-tuning compression protocol based on each file’s distribution, lossless preservation of IUPAC nucleotides and read identifiers, and unrestricted FASTQ/A file attributes (i.e., read length, number of reads, or read identifier format). We compared GeneSqueeze to the general-purpose compressor, gzip, and to a domain-specific compressor, SPRING, to assess performance. Due to GeneSqueeze’s current Python implementation, GeneSqueeze underperformed as compared to gzip and SPRING in the time domain. GeneSqueeze and gzip achieved 100% lossless compression across all elements of the FASTQ files (i.e. the read identifier, sequence, quality score and ‘ + ’ lines). GeneSqueeze and gzip compressed all files losslessly, while both SPRING’s traditional and lossless modes exhibited data loss of non-ACGTN IUPAC nucleotides and of metadata following the ‘ + ’ on the separator line. GeneSqueeze showed up to three times higher compression ratios as compared to gzip, regardless of read length, number of reads, or file size, and had comparable compression ratios to SPRING across a variety of factors. Overall, GeneSqueeze represents a competitive and specialized compression method for FASTQ/A files containing nucleotide sequences. As such, GeneSqueeze has the potential to significantly reduce the storage and transmission costs associated with large omics datasets without sacrificing data integrity.

Improved intensity rounding and division near lossless image compression algorithm using delta encoding

This paper presents RIFD-DLT, an advanced near-lossless image compression algorithm that combines Delta Encoding with the original rounding the intensity followed by division (RIFD) method. The RIFD method first minimizes the image intensities, which makes the next compression stages more efficient. Subsequently, Delta Encoding subtracts neighboring rows in each of the image's three-color matrices, using the proximity of pixel values in adjacent rows to further reduce the image intensity. Extensive investigations show that RIFD-DLT outperforms the state-of-the-art algorithms and benchmarks with respect to compression ratio and processing time. More specifically, RIFD-DLT compresses data from 11520 KBs to 2131 KBs with an 81.93% reduction and a 43.7% improvement over original RIFD-Huffman when compressing Kodak Image set. When comparing the RIFD-DLT with LICA algorithm, the total file size is reduced by 71.2%, representing a 10.73% improvement for RIFD-DLT. Also, RIFD-DLT shows notable speed gains over RIFD-Huffman, requiring only 34.62 seconds to compress and decompress all images from three datasets (Waterloo, Kodak and EPFL), as compared to 50.99 seconds for RIFD-Huffman. As for the image quality, the proposed algorithm achieved an average PSNR values of 58.51 dB, 51.3 dB, and 52.22 dB for the EPFL, Kodak, and Waterloo image sets, respectively, demonstrate the excellent image quality that persists after decompression, with a minimal distortion that is imperceptible to the human visual system and identically to the RIFD-Huffman PSNR. These findings show that, while preserving excellent image quality, RIFD-DLT provides an incredibly efficient and fast method of image compression.

NeutronSketch: An in-depth exploration of redundancy in large-scale graph neural network training

Graph Neural Networks (GNNs) have achieved notable success in various applications. However, the increasing scale of real-world graphs poses a challenge for efficient GNN training. Recent works propose reducing computations by reducing the scale of the input graph through graph compression or sparsification. However, these methods introduce deep learning models requiring multiple training iterations, which leads to significant additional time overhead. Meanwhile, they fail to consider redundant information that is unhelpful or even harmful for model training in the input graphs. In this work, we propose a universal, one-time redundancy removal method called NeutronSketch to remove the redundant information from the input graph. This method can improve training efficiency while maintaining the model accuracy. In the experiments, we compare NeutronSketch with graph compression, sparsification, and coarsening methods. The results show that NeutronSketch has a faster execution speed and better model accuracy. Additionally, we apply NeutronSketch to the sample-based GNN models. The results show that NeutronSketch reduces input graph scale by an average of 25% compared to the original graph, reducing training time by 10%–90% while maintaining the model accuracy.

A comparison between ShapeFit compression and Full-Modelling method with PyBird for DESI 2024 and beyond

DESI aims to provide one of the tightest constraints on cosmological parameters by analysing the clustering of more than thirty million galaxies. However, obtaining such constraints requires special care in validating the methodology and efforts to reduce the computational time required through data compression and emulation techniques. In this work, we perform a rigorous validation of the PyBird power spectrum modelling code with both a traditional emulated Full-Modelling approach and the model-independent ShapeFit compression approach. By using cubic box simulations that accurately reproduce the clustering and precision of the DESI survey, we find that the cosmological constraints from ShapeFit and Full-Modelling are consistent with each other at the ∼ 0.5σ level for the ΛCDM model. Both ShapeFit and Full-Modelling are also consistent with the true ΛCDM simulation cosmology down to a scale of k max = 0.20 hMpc-1 even after including the hexadecapole. For extended models such as the wCDM and the oCDM models, we find that including the hexadecapole can significantly improve the constraints and reduce the modelling errors with the same k max. While their discrepancies between the constraints from ShapeFit and Full-Modelling are more significant than ΛCDM, they remain consistent within 0.7σ. Lastly, we also show that the constraints on cosmological parameters with the correlation function evaluated from PyBird down to s min = 30h -1Mpc are unbiased and consistent with the constraints from the power spectrum.

Adaptive detached eddy simulations of incident shock-induced separation

The low-frequency unsteadiness associated with shock wave turbulent boundary layer interactions (STBLI) poses significant numerical challenges for Reynolds–averaged Navier–Stokes models as well as scale-resolving methods such as direct numerical simulations and large-eddy simulations and requires hybrid methods like detached eddy simulations (DES) from an engineering perspective. The recently developed adaptive DES (ADES) method [Yin and Durbin, “An adaptive DES model that allows wall-resolved eddy simulation,” Int. J. Heat Fluid Flow 62, 499 (2016)], which was originally introduced for low-speed flows and extended to transonic flows, is a promising approach. However, its efficacy for high-speed flows needs to be assessed. In the present study, a compressible version of the ADES method has been implemented and set to test for supersonic flows with complex STBLI features. Impinging STBLI at M∞=2.30 is investigated numerically using compressible ADES method. No inflow turbulence fluctuations are provided in the upstream (time-averaged) boundary layer profile so as to verify the ADES method's capability to predict the low-frequency “breathing” motion of the bubble and the shear layer flapping/shedding in the mid-frequency range, thus demonstrating the significance of downstream mechanism in the unsteadiness associated with STBLI. Time series analysis of surface pressure distribution and modal analysis of numerical schlieren using proper orthogonal decomposition are carried out to understand the dominant mechanisms and associated frequencies. In the interaction zone, starting from the intermittent region to the region downstream of reattachment, a transition of the dominant frequency from low-to-mid-to-high ranges is observed. The separation shock motion and the shear layer shedding are observed to have dominant Strouhal numbers [Stδ0, defined based on the boundary layer thickness (δ0) upstream of interaction] in the ranges of 0.01 and 0.1, respectively, consistent with the literature. From correlation and coherence analyses, it is also observed that the separation bubble is exercising breathing motions at lower frequencies [Stδ0∼O(0.01)] despite lacking the inflow turbulence, thus supporting the downstream influence arguments. The present study demonstrates that the ADES model is capable of predicting essential features of STBLI, including low- and mid-frequency events and a part of high-frequency phenomena.

Coupling immersed boundary method with wall function for high Reynolds number compressible flows on adaptive Cartesian grids

In this paper, a high Reynolds number compressible flow simulation method based on the immersed boundary method (IBM) is developed on the adaptive Cartesian grids to address two problems: one is the ability of large-scale Cartesian grid generation, and the other is the resolution of thin shear layers. For the former, an efficient automatic parallel generation method of adaptive Cartesian grids is proposed based on the k-d tree theory. The method has good parallel scalability and computational efficiency, e.g., the generation of 1.41 × 109 cells by 2048 cores takes only 0.40 min. For the latter, an IBM-wall function coupling method is designed to simulate geometrical configurations with high Reynolds numbers in the parallel framework. Numerical experiments show that the presented method can yield results consistent with the use of body-fitted grids. In conclusion, the presented method can realize a fully automatic simulation of high Reynolds number compressible flows, significantly reducing the labor cost in the computational fluid dynamics process, and can improve the accuracy and efficiency of engineering applications.

Developmental Plasticity-inspired Adaptive Pruning for Deep Spiking and Artificial Neural Networks.

Developmental plasticity plays a prominent role in shaping the brain's structure during ongoing learning in response to dynamically changing environments. However, the existing network compression methods for deep artificial neural networks (ANNs) and spiking neural networks (SNNs) draw little inspiration from brain's developmental plasticity mechanisms, thus limiting their ability to learn efficiently, rapidly, and accurately. This paper proposed a developmental plasticity-inspired adaptive pruning (DPAP) method, with inspiration from the adaptive developmental pruning of dendritic spines, synapses, and neurons according to the "use it or lose it, gradually decay" principle. The proposed DPAP model considers multiple biologically realistic mechanisms (such as dendritic spine dynamic plasticity, activitydependent neural spiking trace, and local synaptic plasticity), with additional adaptive pruning strategy, so that the network structure can be dynamically optimized during learning without any pre-training and retraining. Extensive comparative experiments show consistent and remarkable performance and speed boost with the extremely compressed networks on a diverse set of benchmark tasks for deep ANNs and SNNs, especially the spatio-temporal joint pruning of SNNs in neuromorphic datasets. This work explores how developmental plasticity enables complex deep networks to gradually evolve into brain-like efficient and compact structures, eventually achieving state-of-the-art (SOTA) performance for biologically realistic SNNs.