Size Of Set Research Articles

GOAT: efficient and robust identification of gene set enrichment

Gene set enrichment analysis is foundational to the interpretation of high throughput biology. Identifying enriched Gene Ontology (GO) terms or disease-associated gene sets within a list of gene effect sizes that represent experimental outcomes is an everyday task in life science that crucially depends on robust and sensitive statistical tools. We here present GOAT, a parameter-free algorithm for gene set enrichment analysis of preranked gene lists. The algorithm can precompute null distributions from standardized gene scores, enabling enrichment testing of the GO database in one second. Validations using synthetic data show that estimated gene set p-values are well calibrated under the null hypothesis and invariant to gene list length and gene set size. Application to various real-world proteomics and gene expression studies demonstrates that GOAT identifies more significant GO terms as compared to current methods. GOAT is freely available as an R package and user-friendly online tool for gene set enrichment analyses that includes interactive data visualizations: https://ftwkoopmans.github.io/goat.

Application of a Machine Learning-Based Classification Approach for Developing Host Protein Diagnostic Models for Infectious Disease

In recent years, infectious disease diagnosis has increasingly turned to host-centered approaches as a complement to pathogen-directed ones. The former, however, typically requires the interpretation of complex multiple biomarker datasets to arrive at an informative diagnostic outcome. This report describes a machine learning (ML)-based classification workflow that is intended as a template for researchers seeking to apply ML approaches for developing host-based infectious disease biomarker classifiers. As an example, we built a classification model that could accurately distinguish between three disease etiology classes: bacterial, viral, and normal in human sera using host protein biomarkers of known diagnostic utility. After collecting protein data from known disease samples, we trained a series of increasingly complex Auto-ML models until arriving at an optimized classifier that could differentiate viral, bacterial, and non-disease samples. Even when limited to a relatively small training set size, the model had robust diagnostic characteristics and performed well when faced with a blinded sample set. We present here a flexible approach for applying an Auto-ML-based workflow for the identification of host biomarker classifiers with diagnostic utility for infectious disease, and which can readily be adapted for multiple biomarker classes and disease states.

A meta-analysis on the effects of marker coverage, status number, and size of training set on predictive accuracy and heritability estimates from genomic selection in tree breeding

Genomic selection (GS) is increasingly used in tree breeding because of the possibility to hasten breeding cycles, increase selection intensity or facilitate multi-trait selection, and to obtain less biased estimates of quantitative genetic parameters such as heritability. However, tree breeders are aiming to obtain accurate estimates of such parameters and breeding values while optimizing sampling and genotyping costs. We conducted a metadata analysis of results from 28 GS studies totalling 115 study-traits. We found that heritability estimates obtained using DNA marker-based information for a variety of traits and species were not significantly related to variation in the total number of markers ranging from about 1500 to 116 000, nor by the marker density, ranging from about 1 to 60 markers/centimorgan, nor by the status number of the breeding populations ranging from about 10 to 620, nor by the size of the training set ranging from 236 to 2458. However, the predictive accuracy of breeding values was generally higher when the status number of the breeding population was smaller, which was expected given the higher level of relatedness in small breeding populations, and the increased ability of a given number of markers to trace the long-range linkage disequilibrium in such conditions. According to expectations, the predictive accuracy also increased with the size of the training set used to build marker-based models. Genotyping arrays with a few to many thousand markers exist for several tree species and with the actual costs, GS could thus be efficiently implemented in many more tree breeding programs, delivering less biased genetic parameters and more accurate estimates of breeding values.

An adaptive dual-strategy constrained optimization-based coevolutionary optimizer for high-dimensional feature selection

The feature subset obtained by traditional feature selection algorithms usually contains many irrelevant features and redundant features, which increases the size of the feature set and reduces classification accuracy. More importantly, these irrelevant features and redundant features may lead to local optimum. How to efficiently select the global optimal feature subset from high-dimensional data is a challenging task. To solve the problem, an adaptive dual-strategy constrained optimization-based coevolutionary optimizer for high-dimensional feature selection (ADSCO) is proposed in this paper. The contributions of this algorithm are as follows. Firstly, an adaptive dual strategy constrained optimization mechanism (ADSC) is developed. In the ADSC, the adaptive linear constraint strategy and the random constraint strategy are designed to search for the feature subsets that satisfy the constraints. This can reduce the search space of candidate solutions. Secondly, the dual population-constrained coevolutionary optimizer (CO) is developed to help the algorithm jump out of the local optimum and search for better optimal solutions to further improve the classification accuracy of ADSCO. Finally, the ablation experiments are executed to validate the effectiveness of the proposed ADSC and CO. The ADSCO algorithm is tested on 8 high-dimensional datasets and 8 low-dimensional datasets, and compared with seven typical population feature selection algorithms. Experimental results show that the proposed ADSCO algorithm is superior to the 7 comparison algorithms in terms of classification accuracy and feature subset size, demonstrating its strong competitiveness in tackling high-dimensional data feature selection.

Collaborative learning from distributed data with differentially private synthetic data

BackgroundConsider a setting where multiple parties holding sensitive data aim to collaboratively learn population level statistics, but pooling the sensitive data sets is not possible due to privacy concerns and parties are unable to engage in centrally coordinated joint computation. We study the feasibility of combining privacy preserving synthetic data sets in place of the original data for collaborative learning on real-world health data from the UK Biobank.MethodsWe perform an empirical evaluation based on an existing prospective cohort study from the literature. Multiple parties were simulated by splitting the UK Biobank cohort along assessment centers, for which we generate synthetic data using differentially private generative modelling techniques. We then apply the original study’s Poisson regression analysis on the combined synthetic data sets and evaluate the effects of 1) the size of local data set, 2) the number of participating parties, and 3) local shifts in distributions, on the obtained likelihood scores.ResultsWe discover that parties engaging in the collaborative learning via shared synthetic data obtain more accurate estimates of the regression parameters compared to using only their local data. This finding extends to the difficult case of small heterogeneous data sets. Furthermore, the more parties participate, the larger and more consistent the improvements become up to a certain limit. Finally, we find that data sharing can especially help parties whose data contain underrepresented groups to perform better-adjusted analysis for said groups.ConclusionsBased on our results we conclude that sharing of synthetic data is a viable method for enabling learning from sensitive data without violating privacy constraints even if individual data sets are small or do not represent the overall population well. Lack of access to distributed sensitive data is often a bottleneck in biomedical research, which our study shows can be alleviated with privacy-preserving collaborative learning methods.

Conformalized prescriptive machine learning for uncertainty-aware automated decision making: the case of goodwill requests

AbstractDue to the inherent presence of uncertainty in machine learning (ML) systems, the usage of ML is until now out of scope for many critical (financial) business processes. One such process is goodwill assessment at car manufacturers, where a large part of goodwill cases is still assessed manually by human experts. To increase the degree of automation while still providing an overall reliable assessment service, we propose a selective uncertainty-aware automated decision making approach based on uncertainty quantification through conformal prediction. In our approach, goodwill requests are still shifted to human experts in case the risk of a wrong assessment is too high. Nevertheless, ML can be introduced into the process with reduced and controllable risk. We hereby determine the risk of wrong ML assessments through two hierarchical conformal predictors that make use of the prediction set and interval size as the main criteria for quantifying uncertainty. We also utilize conformal prediction’s property to output empty prediction sets if no prediction is significant enough and abstain from an automatic decision in that case. Instead of providing mathematical guarantees for limited risk, we focus on the risk vs. degree of automation trade-off and how a business decision maker can select in an a posteriori fashion a trade-off that best suits the business problem at hand from a set of pareto optimal solutions. We also show empirically on a goodwill data set of a BMW National Sales Company that by only selecting certain requests for automated decision making we can significantly increase the accuracy of automatically processed requests. For instance, from 92 to 98% for labor and from 90 to 98% for parts contributions respectively, while still maintaining a degree of automation of approximately 70%.

Enhancing IoT security: A comparative study of feature reduction techniques for intrusion detection system

Internet of Things (IoT) devices are extensively utilized but are susceptible to cyberattacks, posing significant security challenges. To mitigate these threats, machine learning techniques have been implemented for network intrusion detection in IoT environments. These techniques commonly employ various feature reduction methods, prior to inputting data into models, in order to enhance the efficiency of detection processes to meet real-time requirements. This study provides a comprehensive comparison of feature selection (FS) and feature extraction (FE) techniques for network intrusion detection systems (NIDS) in IoT environments, utilizing the TON-IoT and BoT-IoT datasets for both binary and multi-class classification tasks. We evaluated FS methods, including Pearson correlation and Chi-square, and FE methods, such as Principal Component Analysis (PCA) and Autoencoders (AE), across five classic machine learning models: Decision Tree (DT), Random Forest (RF), Naive Bayes (NB), k-Nearest Neighbors (kNN), and Multi-Layer Perceptron (MLP). Our analysis revealed that FE techniques generally achieve higher accuracy and robustness compared to FS methods, with RF paired with AE delivering superior performance despite higher computational demands. DTs are most effective with smaller feature sets, while MLPs excel with larger sets. Chi-square is identified as the most efficient FS method, balancing performance and computational efficiency, whereas PCA outperforms AE in runtime efficiency. The study also highlights that FE methods are more effective for complex datasets and less sensitive to feature set size, whereas FS methods show significant performance improvements with more informative features. Despite the higher computational costs of FE methods, they demonstrate a greater capability to detect diverse attack types, making them particularly suitable for complex IoT environments. These findings are crucial for both academic research and industry applications, providing insights into optimizing detection performance and computational efficiency in NIDS for IoT networks.

Three-way decision based island harmony search algorithm for robust flow-shop scheduling with uncertain processing times depicted by big data

This paper discusses an uncertain two-machine permutation flow-shop scheduling problem (2PFSP) with total weighted tardiness and common due date. Uncertain processing times are described by a large set of discrete scenarios, which is a type of big data. The objective is to minimize the schedule performance under the worst-case scenario. Identifying the worst-case scenario for each evaluated schedule is quite time-consuming in the situation that the scenario set size is large so that the objective evaluation might be computationally expensive. To handle this difficulty, three-way decision is used to preprocess the large-size scenario set to get a reduced scenario set so that the concept of surrogate worst-case scenario is adopted. A hybrid harmony search algorithm of combining three-island framework and the scenario-based local search is developed to solve the discussed problem. Based on the single-scenario knowledge of 2PFSP, a problem-specific scenario-dependent neighborhood structure is constructed under the surrogate worst-case scenario. An extensive experiment was carried out. The computational results show that the application of surrogate worst-case scenario based on three-way decision is effective in reducing the time consuming while keeping schedule performance evaluation. Being compared to the worst-case scenario objective evaluation, for an example in the case of the middle bad-scenario ratio, the surrogate worst-case scenario objective evaluation made the solution algorithm save 12.95 % in average CPU time for all instances while the relative performance difference is only 1.809 % in average. Being compared to possible alternative algorithms derived from the state-of-the-art algorithms, the developed algorithm is advantageous for the addressed problems.

Layer-by-layer reconstruction of fatigue damages in composites from thermal images by a Residual U-Net

In this paper a deep learning model is used to fully reconstruct the 3D distribution of arbitrarily shaped subsurface fatigue damages in a fiber/epoxy composite from synthetic thermal surface images. Synthetic thermal surface images (TIs) of self-heating damage hotspots are produced by thermal finite element analysis which are consequently used to train a Residual U-Net based on recent architectures designed for image segmentation. Different augmentation techniques are employed to mitigate the computational cost of generating training data through thermal finite element analysis. The Residual U-Net model accurately reconstructed – layer by layer – the ground truths and thereby enabled the quantitative assessment of location, size, shape, depth and gradient of an internal fatigue damage distribution. Moreover, the Residual U-Net achieved good predictions for comparatively small training set sizes.

On data efficiency of univariate time series anomaly detection models

In machine learning (ML) problems, it is widely believed that more training samples lead to improved predictive accuracy but incur higher computational costs. Consequently, achieving better data efficiency, that is, the trade-off between the size of the training set and the accuracy of the output model, becomes a key problem in ML applications. In this research, we systematically investigate the data efficiency of Univariate Time Series Anomaly Detection (UTS-AD) models. We first experimentally examine the performance of nine popular UTS-AD algorithms as a function of the training sample size on several benchmark datasets. Our findings confirm that most algorithms become more accurate when more training samples are used, whereas the marginal gain for adding more samples gradually decreases. Based on the above observations, we propose a novel framework called FastUTS-AD that achieves improved data efficiency and reduced computational overhead compared to existing UTS-AD models with little loss of accuracy. Specifically, FastUTS-AD is compatible with different UTS-AD models, utilizing a sampling- and scaling law-based heuristic method to automatically determine the number of training samples a UTS-AD model needs to achieve predictive performance close to that when all samples in the training set are used. Comprehensive experimental results show that, for the nine popular UTS-AD algorithms tested, FastUTS-AD reduces the number of training samples and the training time by 91.09–91.49% and 93.49–93.82% on average without significant decreases in accuracy.

A multi-objective optimisation approach with improved pareto-optimal solutions to enhance economic and environmental dispatch in power systems

This work implements the recently developed nth state Markovian jumping particle swarm optimisation (PSO) algorithm with local search (NS-MJPSOloc) awareness method to address the economic/environmental dispatch (EED) problem. The proposed approach, known as the Non-dominated Sorting Multi-objective PSO with Local Best (NS-MJPSOloc), aims to enhance the performance of the PSO algorithm in multi-objective optimisation problems. This is achieved by redefining the concept of best local candidates within the search space of multi-objective optimisation. The NS-MJPSOloc algorithm uses an evolutionary factor-based mechanism to identify the optimum compromise solution, a Markov chain state jumping technique to control the Pareto-optimal set size, and a neighbourhood’s topology (such as a ring or a star) to determine its size. Economic dispatch refers to the systematic allocation of available power resources in order to fulfill all relevant limitations and effectively meet the demand for electricity at the lowest possible operating cost. As a result of heightened public consciousness regarding environmental pollution and the implementation of clean air amendments, nations worldwide have compelled utilities to adapt their operational practises in order to comply with environmental regulations. The (NS-MJPSOloc) approach has been utilised for resolving the EED problem, including cost and emission objectives that are not commensurable. The findings illustrate the efficacy of the suggested (NS-MJPSOloc) approach in producing a collection of Pareto-optimal solutions that are evenly dispersed within a single iteration. The comparison of several approaches reveals the higher performance of the suggested (NS-MJPSOloc) in terms of the diversity of the Pareto-optimal solutions achieved. In addition, a measure of solution quality based on Pareto optimality has been incorporated. The findings validate the effectiveness of the proposed (NS-MJPSOloc) approach in addressing the multi-objective EED issue and generating a trade-off solution that is both optimal and of high quality. We observed that our approach can reduce ∼\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$\\sim $$\\end{document}6.4% of fuel costs and ∼\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$\\sim $$\\end{document}9.1% of computational time in comparison to the classical PSO technique. Furthermore, our method can reduce ∼\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$\\sim $$\\end{document}9.4% of the emissions measured in tons per hour as compared to the PSO approach.

MLC-miner: Efficiently discovering multi-level closed high utility patterns from quantitative hierarchical transaction databases

High utility pattern mining (HUPM) extends frequent pattern mining (FPM) by including item significance and quantity, which determine their utility, in transaction databases. A pattern is a high utility if its utility satisfies the minimum threshold. Several efficient approaches have been developed to address this task. However, more extensive databases combined with lower thresholds can lead to an explosion in the number of discovered patterns. Use the closed version of the pattern to decrease patterns while keeping information. To improve real-world applicability, HUPM research now considers item connections. Categorizing items creates an abstraction model that can help decision-makers. This new mining task is generalized high utility pattern mining (GHUPM). However, GHUPM significantly increases the complexity of the mining task, resulting in longer mining times, higher memory usage, and a more significant number of discovered patterns. To date, none of the proposed approaches in GHUPM have utilized the closed representation to shrink the size of the result set. This study introduces MLC-Miner, a multi-level closed high utility pattern extraction technique, to fill this research gap. MLC-Miner combines several efficient methods to enhance mining performance and reduce memory footprints. Experimental results show that MLC-Miner can handle up to 5 million transactions efficiently.

Working memory limitations constrain visual episodic long-term memory at both specific and gist levels of representation.

Limitations in one's capacity to encode information in working memory (WM) constrain later access to that information in long-term memory (LTM). The present study examined whether these WM constraints on episodic LTM are limited to specific representations of past episodes or also extend to gist representations. Across three experiments, young adult participants (n = 40 per experiment) studied objects in set sizes of two or six items, either sequentially (Experiments 1a and 1b) or simultaneously (Experiment 2). They then completed old/new recognition tests immediately after each sequence (WM tests). After a long study phase, participants completed LTM conjoint recognition tests, featuring old but untested items from the WM phase, lures that were similar to studied items at gist but not specific levels of representation, and new items unrelated to studied items at both specific and gist levels of representation. Results showed that LTM estimates of specific and gist memory representations from a multinomial-processing-tree model were reduced for items encoded under supra-capacity set sizes (six items) relative to within-capacity set sizes (two items). These results suggest that WM encoding capacity limitations constrain episodic LTM at both specific and gist levels of representation, at least for visual objects. The ability to retrieve from LTM each type of representation for a visual item is contingent on the degree to which the item could be encoded in WM.

Residue sums of Dickson polynomials over finite fields

Given a polynomial with integral coefficients, one can inquire about the possible residues it can take in its image modulo a prime p. The sum over the distinct residues can sometimes be computed independent of the prime p; for example, Gauss showed that the sum over quadratic residues vanishes modulo a prime. In this paper we provide a closed form for the sum over distinct residues in the image of Dickson polynomials of arbitrary degree over finite fields of odd characteristic, and prove a complete characterization of the size of the value set. Our result provides the first non-trivial classification of such a sum for a family of polynomials of unbounded degree.

Multiple visual items can be simultaneously compared with target templates in memory.

When we search for something, we often rely on both what we see and what we remember. This process can be divided into three stages: selecting items, identifying those items, and comparing them with what we are trying to find in our memory. It has been suggested that we select items one by one, and we can identify several items at once. In the present study, we tested whether we need to finish comparing a selected item in the visual display with one or more target templates in memory before we can move on to the next selected item. In Experiment 1, observers looked for either one or two target types in a rapid serially presented stimuli stream. The time interval between the presentation onset of successive items in the stream was varied to get a threshold. For search for one target, the threshold was 89ms. When look for either of two targets, it was 192ms. This threshold difference offered a baseline. In Experiment 2, observers looked for one or two types of target in a search array. If they compared each identified item separately, we should expect a jump in the slope of the RT × Set Size function, on the order of the baseline obtained in Experiment 1. However, the slope difference was only 13ms/item, suggesting that several identified items can be compared at once with target templates in memory. Experiment 3 showed that this slope difference was not just a memory-load cost.

Sparsity-Preserved Pareto Optimization for Subset Selection

Subset selection that selects a limited number of variables optimizing many given criteria, is a fundamental problem with various applications such as sparse regression and unsupervised feature selection. Among the existing algorithms for subset selection, evolutionary algorithms (EAs) have achieved great performance in obtaining the subsets with high quality. However, as the selected subsets are sparse, i.e., most variables of these solutions are zero, most existing EAs for subset selection problems pose great challenges to find the optimal solutions effectively and efficiently, especially when the size of original set is large. To this end, we propose the SPESS approach which is a sparsity preserved evolutionary algorithm for subset selection. To be specific, we design two sparsity preserved operators (i.e., sparsity preserved crossover operator and sparsity preserved mutation operator), which can be used in SPESS to ensure the sparsity of the generated solutions in the process of Pareto optimization. The main advantages of the proposed sparsity preserved operators are that they can obtain high quality solutions and improve the search efficiency of the algorithm. According to the experimental results on 12 sparse regression problems and 12 unsupervised feature selection problems, the proposed SPESS is superior over the state-of-the-arts in solving subset selection tasks and the effectiveness of the proposed two sparsity preserved operators is also verified.

A Comparison of CT-Based Pancreatic Segmentation Deep Learning Models

Rationale and ObjectivesPancreas segmentation accuracy at CT is critical for the identification of pancreatic pathologies and is essential for the development of imaging biomarkers. Our objective was to benchmark the performance of five high-performing pancreas segmentation models across multiple metrics stratified by scan and patient/pancreatic characteristics that may affect segmentation performance. Materials and MethodsIn this retrospective study, PubMed and ArXiv searches were conducted to identify pancreas segmentation models which were then evaluated on a set of annotated imaging datasets. Results (Dice score, Hausdorff distance [HD], average surface distance [ASD]) were stratified by contrast status and quartiles of peri-pancreatic attenuation (5mm region around pancreas). Multivariate regression was performed to identify imaging characteristics and biomarkers (n=9) that were significantly associated with Dice score. ResultsFive pancreas segmentation models were identified: Abdomen Atlas [AAUNet, AASwin, trained on 8448 scans], TotalSegmentator [TS, 1204 scans], nnUNetv1 [MSD-nnUNet, 282 scans], and a U-Net based model for predicting diabetes [DM-UNet, 427 scans]. These were evaluated on 352 CT scans (30 females, 25 males, age 58±7 years old [±1SD], 327 age unknown) from 2000-2023. Overall, TS, AAUNet, and AASwin were the best performers, Dice=80±11%, 79±16%, and 77±18% respectively (pairwise Sidak test not-significantly different). AASwin and MSD-nnUNet performed worse (for all metrics) on non-contrast scans (vs contrast, P<.001). The worst performer was DM-UNet (Dice=67±16%). All algorithms except TS showed lower Dice scores with increasing peri-pancreatic attenuation (P<.01). Multivariate regression showed non-contrast scans, (P<.001; MSD-nnUNet), smaller pancreatic length (P=.005, MSD-nnUNet), and height (P=.003, DM-UNet) were associated with lower Dice scores. ConclusionThe convolutional neural network-based models trained on a diverse set of scans performed best (TS, AAUnet, and AASwin). TS performed equivalently to AAUnet and AASwin with only 13% of the training set size (8488 vs 1204 scans). Though trained on the same dataset, a transformer network (AASwin) had poorer performance on non-contrast scans whereas its convolutional network counterpart (AAUNet) did not. This study highlights how aggregate assessment metrics of pancreatic segmentation algorithms seen in other literature are not enough to capture differential performance across common patient and scanning characteristics in clinical populations.

How many specimens make a sufficient training set for automated three-dimensional feature extraction?

Deep learning has emerged as a robust tool for automating feature extraction from three-dimensional images, offering an efficient alternative to labour-intensive and potentially biased manual image segmentation methods. However, there has been limited exploration into the optimal training set sizes, including assessing whether artficial expansion by data augmentation can achieve consistent results in less time and how consistent these benefits are across different types of traits. In this study, we manually segmented 50 planktonic foraminifera specimens from the genus Menardella to determine the minimum number of training images required to produce accurate volumetric and shape data from internal and external structures. The results reveal unsurprisingly that deep learning models improve with a larger number of training images with eight specimens being required to achieve 95% accuracy. Furthermore, data augmentation can enhance network accuracy by up to 8.0%. Notably, predicting both volumetric and shape measurements for the internal structure poses a greater challenge compared with the external structure, owing to low contrast differences between different materials and increased geometric complexity. These results provide novel insight into optimal training set sizes for precise image segmentation of diverse traits and highlight the potential of data augmentation for enhancing multivariate feature extraction from three-dimensional images.

A Hybrid Neural Network-Based Fast Financial Fraud Detection Model

With the increasing number of financial transactions, financial fraud has become increasingly serious for financial institutions and the public. The core idea of this model is to integrate multiple neural network structures and utilize their respective advantages to improve the performance of fraud detection. Firstly, we employed the convolutional neural network with interpretable blocks (CNNIB) convolutional neural network (CNN) to extract key features from the data to capture patterns and patterns in fraud cases. Secondly, we introduced the autoencoder generative adversarial network (AE-GAN) adversarial network to perform feature analysis on sequence data to capture temporal features in transaction sequences. Finally, we used differential detection for classification to determine whether transactions were fraudulent. An independent detection module was established to accelerate the recognition of financial fraud, and parameter indicators were optimized. Finally, a hybrid neural network model was established. The experimental results indicate that our model has achieved significant results in quickly detecting financial fraud; compared with traditional single neural network models, hybrid neural network models have significant improvements in accuracy and efficiency. In addition, we conducted in-depth analysis of the model and revealed its performance stability under different training set sizes and data distributions. Our research findings provide an effective tool for financial institutions to quickly identify financial fraud.

Stagewise Training With Exponentially Growing Training Sets.

In the world of big data, training large-scale machine learning problems has gained considerable attention. Numerous innovative optimization strategies have been presented in recent years to accelerate the large-scale training process. However, the possibility of further accelerating the training process of various optimization algorithms remains an unresolved subject. To begin addressing this difficult problem, we exploit the researched findings that when training data are independent and identically distributed, the learning problem on a smaller dataset is not significantly different from the original one. Upon that, we propose a stagewise training technique that grows the size of the training set exponentially while solving nonsmooth subproblem. We demonstrate that our stagewise training via exponentially growing the size of the training sets (STEGSs) are compatible with a large number of proximal gradient descent and gradient hard thresholding (GHT) techniques. Interestingly, we demonstrate that STEGS can greatly reduce overall complexity while maintaining statistical accuracy or even surpassing the intrinsic error introduced by GHT approaches. In addition, we analyze the effect of the training data growth rate on the overall complexity. The practical results of applying l2,1 -and l0 -norms to a variety of large-scale real-world datasets not only corroborate our theories but also demonstrate the benefits of our STEGS framework.