Pith. sign in

REVIEW 4 major objections 6 minor 87 references

MamNet: A Novel Hybrid Model for Time-Series Forecasting and Frequency Pattern Analysis in Network Traffic

T0 review · 4 major / 6 minor · reviewed 2026-08-06 · deepseek-v4-flash

Pith's one-line read The paper claims that combining Mamba's state-space memory with Fourier frequency features improves network traffic anomaly detection by about 2–4% over five recent baselines.

desk verdict This paper's central 2–4% improvement claim is unverifiable as submitted: the evaluation protocol is missing and the main results table is corrupted. read the letter →

arxiv 2507.00304 v1 pith:CTJCBG5N submitted 2025-06-30 cs.LG cs.NI

classification cs.LGcs.NI
keywords networktrafficpredictionanomalydetectionMambastate-spacemodelFourierTransformtime-frequencyfusionlong-termdependencyperiodicity
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

MamNet is an encoder–decoder model that fuses a Mamba state-space module for time-domain modeling with a Fourier Transform module for frequency-domain feature extraction, using learnable weights to combine the two. The paper tries to establish that this hybrid captures both long-term dependencies and periodic fluctuations in network traffic better than recent time-series models, and that the gain is concrete: roughly 2–4% higher accuracy, recall, and F1-Score plus 10–25% lower MAE and MSE on the UNSW-NB15 and CAIDA datasets. If correct, network operators would gain an anomaly detector that catches scheduled peaks and attack patterns while using less memory and lower inference latency than LSTM and GRU baselines.

What carries the argument

The engine of the argument is the weighted fusion $z_t = \alpha x_t + \beta X(f)$, in which $x_t$ is Mamba's time-domain output and $X(f)$ is the Fourier Transform of the traffic signal, with $\alpha$ and $\beta$ learnable through backpropagation. Mamba is a state-space model whose update $x_{t+1} = A x_t + B u_t$, $y_t = C x_t + D u_t$ gives linear-complexity long-range memory, and the Fourier Transform exposes the periodic components that time-domain models miss. The feature fusion layer combines these two representations at multiple scales so the model can weight long-term trend information against periodic-pattern information per scenario; this pairing is what the paper credits for the reported 2–4% gains.

What would settle it

Shuffle the rows within each dataset to destroy temporal order while keeping every feature and label, then retrain MamNet under the same protocol; if accuracy, recall, and F1-Score stay near the reported 96% and 94%, the long-term-dependency and periodicity components are not producing the claimed gains.

Watch

Extended reading notes

Core claim

On the paper's own terms, the central discovery is that the combination of time-domain and frequency-domain branches in MamNet yields consistently better traffic prediction and anomaly detection than using either branch alone. On UNSW-NB15, MamNet reports 96.32% accuracy, 95.16% recall, and 95.74% F1-Score with MAE 0.022 and MSE 0.016; on CAIDA it reports 94.78% accuracy, 92.42% recall, and 93.56% F1-Score with MAE 0.031 and MSE 0.021. Against Temporal Fusion Transformer, Informer, N-BEATS, Autoformer, and FlowForecaster, the paper reports consistent gains: roughly 2–3% in accuracy and F1-Score on UNSW-NB15, about 2% in recall on CAIDA, and 20–25% or 10–15% reductions in MAE and MSE depending on dataset. Ablation results show that removing the Mamba module costs about 1–2% accuracy, removing the Fourier module costs about 1–2% recall and F1-Score, and removing both costs 2–3%, which the paper reads as evidence that the two modules are complementary.

Load-bearing premise

The load-bearing premise is that UNSW-NB15 and CAIDA are actually being used as chronological time series with a defined sequence length and forecast horizon; the paper never specifies sequence construction, train/test split, or how labels map to time windows, so if the rows are shuffled flow records, the claimed temporal and frequency effects are not what the experiments measure.

Editorial extensions

If this is right

  • If MamNet's results hold, the fused model outperforms TFT, Informer, N-BEATS, Autoformer, and FlowForecaster on both datasets across accuracy, recall, F1-Score, MAE, and MSE.
  • Each module is load-bearing: ablations put the cost of removing Mamba at about 1–2% accuracy and the cost of removing the Fourier branch at about 1–2% recall and F1-Score, with errors rising when both are removed.
  • MamNet's reported efficiency (16 GB memory, about 20 ms inference latency, 10 hours of training on a data-center GPU) makes real-time deployment feasible despite the extra Fourier branch.
  • In dynamic-topology tests, MamNet's accuracy and recall drop roughly 2% and 1.5%, versus 5–6% for TFT and Informer, suggesting the frequency branch keeps periodicity detection stable when the network graph changes.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • A testable extension is to rank attack types by their spectral peak strength and check whether MamNet's recall gain over the baselines grows with periodicity; the paper's mechanism predicts it should, but the paper does not report such a stratification.
  • The efficiency comparison is made against LSTM and GRU, not against other time-frequency hybrids; comparing with Fourier-augmented transformers would isolate what the frequency branch actually costs.
  • If the mechanism transfers, the same time-domain-plus-frequency fusion recipe could be applied to other strongly seasonal forecasting tasks, such as electricity load or cloud workload prediction, with Mamba replaced by any efficient sequence encoder.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 6 minor

Summary. The paper proposes MamNet, an encoder-decoder model that combines Mamba-based time-domain state-space modeling with Fourier-based frequency-domain feature extraction, followed by a learnable weighted fusion of the two representations, for network traffic prediction and anomaly detection. The authors claim that on UNSW-NB15 and CAIDA, MamNet outperforms TFT, Informer, N-BEATS, Autoformer, and FlowForecaster by roughly 2-4% in accuracy, recall, and F1-Score while also reducing MAE and MSE, and that ablation experiments confirm the contribution of each module. The central contribution is empirical. As submitted, however, the experimental protocol is not defined: Section 4.1 does not specify how the datasets are converted into ordered time series, what horizon or split is used, or how labels are mapped, and Table 2 is internally inconsistent. The missing equations and underspecified fusion equation make the model definition incomplete. The paper therefore does not, in its current form, provide verifiable support for the stated claims.

Significance. If the empirical claims were fully supported, the contribution would be a modest incremental advance: combining a state-space sequence model with Fourier analysis is a plausible and increasingly common design, and the paper's intended separation of long-term dependency modeling from periodicity extraction is conceptually clear. The manuscript provides no code, no data-release link, and no machine-checked derivations, so reproducibility rests entirely on the textual description. The clearest strength is the explicit ablation design that isolates time-domain, frequency-domain, and fusion components. However, because the evaluation protocol is missing and the results table is corrupted, the central quantitative claim - the 2-4% improvement - cannot currently be assessed. The proposed architecture's significance therefore remains unestablished in this submission.

major comments (4)
  1. [§4.1] The paper never defines a temporal evaluation protocol for UNSW-NB15 and CAIDA. Section 4.1 describes only min-max normalization, Recursive Feature Elimination, and SMOTE-style oversampling/undersampling, all of which are row-level operations. There is no statement of sequence length, forecast horizon, chronological train/validation/test split, or mapping from the datasets' labels to the anomaly-detection task. Without this, the Mamba module's sequential state updates and the Fourier module's frequency extraction cannot be shown to operate on ordered time series; if rows were shuffled and randomly split as tabular records, the reported gains would not measure long-term dependencies or periodic patterns at all. This makes the central claim unverifiable and is the most load-bearing defect in the manuscript.
  2. [Table 2] Table 2 is corrupted and cannot support the paper's statistical claims. In the MamNet CAIDA row, an interval (95.50-97.14) is immediately followed by point values 94.78, 92.42, 93.56, and further intervals, so the reader cannot determine which point estimate corresponds to which confidence interval; the MamNet CAIDA accuracy point estimate is missing or misplaced. Several baseline rows also repeat identical metric values (e.g., the Informer UNSW-NB15 row entries match the TFT UNSW-NB15 row entries for Recall, F1-Score, MAE, and MSE). The text states that t-tests and 95% confidence intervals were computed, but no test statistics, p-values, or number of repetitions are reported. The abstract's 2-4% improvement therefore cannot be matched to any internally consistent result.
  3. [§3.2–§3.3] The model definition is incomplete and internally inconsistent. Equation (3), the Fourier transform, is missing from the text, as are Equations (5), (7), and the equation numbers for the metrics in Section 4.2. Equation (4) defines z_t = α·x_t + β·X(f), mixing a time-domain state or vector x_t with a frequency-domain spectrum X(f) without specifying the dimensions of either object, the normalization applied to the spectrum, or whether the Fourier transform is computed causally over past windows only. If X(f) is computed over the whole sequence or a centered window, the fused representation at time t may contain future information, and the reported anomaly-detection results would then be affected by leakage. As written, the architecture cannot be reproduced or checked.
  4. [§4.3] The paper makes additional empirical claims that are presented without any protocol or supporting table: per-attack-type results for DDoS, port scanning, and data exfiltration, and experiments under dynamic topology changes with quantitative drops in accuracy, recall, MAE, and MSE. No description is given of how these experiments were constructed, what data they used, or how the reported percentages were measured. These claims are therefore uncheckable and should be removed or fully documented.
minor comments (6)
  1. [References] Reference [40] is cited as FlowForecaster, but the cited work is titled 'Probabilistic load forecasting with generative models' and does not appear to correspond to a model named FlowForecaster. The baseline identity should be corrected or replaced with a proper citation.
  2. [Abstract and §4.2] The task definition is inconsistent: the abstract and the use of MAE/MSE imply continuous traffic-value forecasting, while accuracy, recall, F1-Score, and the phrase 'abnormal traffic' imply binary anomaly classification. Section 4 must state clearly whether the model outputs a scalar forecast, a class label, or both, and how the classification metrics are computed from the model output.
  3. [Table 1] Table 1 lists 'Scanning' as an attack type in CAIDA, which is not standard for the CAIDA DDoS datasets, and reports CAIDA with 40,000 samples; the authors should clarify which exact CAIDA collection was used and how the sample was extracted and labeled.
  4. [Figures] Figures 2, 3, and 5 are referenced in the text but never described; Figure 4 has only the uninformative caption 'Visualization of comparative experiments for reference model advantages' and is not discussed in the body text. Please add proper descriptions and, where relevant, error bars or shaded confidence regions.
  5. [References] The reference list contains many entries that are not cited in the text (e.g., [43]-[87] largely appear uncited) and numerous self-citations with only tangential relevance, such as [56] and [57]. The citation list should be trimmed to works actually used.
  6. [Throughout] There are formatting and typographical issues, including the odd spacing in 'O(n2 )', the typo 'naıve', and the use of '四' before the corresponding author's name. These do not affect the technical content but should be corrected.

Circularity Check

0 steps flagged · score 0.0 of 10

No circular derivation; MamNet is a standard trainable hybrid (SSM + Fourier + learned weighted fusion) compared against external baselines, so the central claim is not equivalent to an input by construction.

full rationale

I walked the derivation chain: Sec. 3.2 defines the time-domain module via the standard SSM equations x_{t+1}=Ax_t+Bu_t, y_t=Cx_t+Du_t (Eqs. 1-2); Sec. 3.3 defines frequency features via Fourier transform and fuses them with learnable weights alpha and beta (Eq. 4) optimized by backpropagation. None of these equations define the claimed 2-4% improvement; the improvement is an empirical comparison in Sec. 4.3 against TFT, Informer, N-BEATS, Autoformer, and FlowForecaster on UNSW-NB15 and CAIDA. The evaluation is not circular: the model is trained and then compared on held-out metrics, and the baselines are external published methods. The reference list contains numerous same-author citations (e.g., refs. 22, 56, 64, 70, 80), but none is used to justify MamNet's architecture or to forbid alternatives; the architecture rests on the standard Mamba/SSM and Fourier-transform literature and on the paper's own learnable components. The serious weaknesses in the paper are the missing equations (3), (5), and (7), the undefined temporal train/test protocol for UNSW-NB15 and CAIDA in Sec. 4.1, and the corrupted Table 2 confidence intervals; these undermine verifiability of the empirical claim, but they are not circularity, because no equation, fitted parameter, or self-citation is shown to be equivalent by construction to the output being predicted.

Assumptions & free parameters 5 free parameters · 4 assumptions · 0 invented entities

The model's evidence burden is almost entirely empirical: the fusion rule, Fourier window, Mamba hyperparameters, and preprocessing choices are free or tuned components. Because none are pinned down by code or exact values, the central comparison cannot be reproduced from the manuscript.

free parameters (5)
  • Fusion weights alpha and beta = not reported
    Equation (4) defines the fused feature as alpha * x_t + beta * X(f); these weights are learned by backpropagation and are central to the claimed multi-scale fusion, but their fitted values are not given.
  • Fourier window size and frequency resolution = not reported
    Section 3.3 states these were chosen by grid search and sensitivity analysis; they directly determine which periodic components are extracted.
  • Mamba hyperparameters = not reported
    Section 3.2 states the state transition matrix initialization, learning rate, and iteration count were selected by grid search and cross-validation; no selected values are listed.
  • Training hyperparameters = learning rate 0.001, batch size 32, dropout 0.3
    Section 3.3 fixes these values; no sensitivity analysis or alternative settings are reported.
  • Feature selection and class-balancing choices = not reported
    Section 4.1 applies Recursive Feature Elimination, correlation filtering, SMOTE, and undersampling, but the final feature count, thresholds, and sampling ratios are not stated, altering the evaluated data distribution.
assumptions (4)
  • domain assumption The UNSW-NB15 and CAIDA records are chronologically ordered time series whose temporal structure is the object of prediction.
    Section 4.1 lists datasets and preprocessing but does not define sequence length, horizon, split, or label construction; without this, long-term dependency and periodicity claims are not measurable.
  • domain assumption Fourier transform of a windowed traffic signal exposes the periodic patterns relevant to anomaly detection.
    Section 3.3 assumes periodic attacks and scheduled peaks appear as stable frequency components in the chosen window.
  • ad hoc to paper A scalar weighted sum of time-domain and frequency-domain features is sufficient for multi-scale fusion.
    Equation (4) adopts this fusion rule; Section 4.4 mentions alternative concatenation and averaging only qualitatively, with no comparison table.
  • standard math The Mamba state-space equations (1)-(2) and the O(n) complexity claim describe the implemented module.
    The paper relies on the standard Mamba/SSM formulation from prior literature instead of deriving or validating the implementation here.

how reviews work

0 comments
Cite this review

Pith. "Pith review of MamNet: A Novel Hybrid Model for Time-Series Forecasting and Frequency Pattern Analysis in Network Traffic." pith.science (2026). https://pith.science/paper/CTJCBG5N

@misc{pith2026250700304,
  author       = {Pith},
  title        = {Pith review of: MamNet: A Novel Hybrid Model for Time-Series Forecasting and Frequency Pattern Analysis in Network Traffic},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/CTJCBG5N}},
  note         = {Machine review of arXiv:2507.00304}
}
read the original abstract

The abnormal fluctuations in network traffic may indicate potential security threats or system failures. Therefore, efficient network traffic prediction and anomaly detection methods are crucial for network security and traffic management. This paper proposes a novel network traffic prediction and anomaly detection model, MamNet, which integrates time-domain modeling and frequency-domain feature extraction. The model first captures the long-term dependencies of network traffic through the Mamba module (time-domain modeling), and then identifies periodic fluctuations in the traffic using Fourier Transform (frequency-domain feature extraction). In the feature fusion layer, multi-scale information is integrated to enhance the model's ability to detect network traffic anomalies. Experiments conducted on the UNSW-NB15 and CAIDA datasets demonstrate that MamNet outperforms several recent mainstream models in terms of accuracy, recall, and F1-Score. Specifically, it achieves an improvement of approximately 2% to 4% in detection performance for complex traffic patterns and long-term trend detection. The results indicate that MamNet effectively captures anomalies in network traffic across different time scales and is suitable for anomaly detection tasks in network security and traffic management. Future work could further optimize the model structure by incorporating external network event information, thereby improving the model's adaptability and stability in complex network environments.

Figures

Figures reproduced from arXiv: 2507.00304 by the authors.

Figure 2
Figure 2. Mamba Model Time-Domain Modeling Architecture. The working principle of the Mamba model is based on modeling dynamic systems, where network traffic data is treated as a linear dynamic system. The model uses hidden states to capture long-term dependencies in the data. Specifically, the state-space model describes the dynamic behavior of the system through a set of state equations. The state transition equation descri… view at source ↗
Figure 4
Figure 4. Visualization of comparative experiments for reference model advantages [PITH_FULL_IMAGE:figures/full_fig_p011_4.png] view at source ↗

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

87 extracted references · 77 canonical work pages

  1. [1]

    Deep Learning on Network Traffic Prediction: Recent Advances, Analysis, and Future Directions

    Ons Aouedi et al. “ Deep Learning on Network Traffic Prediction: Recent Advances, Analysis, and Future Directions”. In: ACM Computing Surveys 57.6 (2025), pp. 1–37

  2. [2]

    Short -term OD flow predicti on for urban rail transit control: A multi -graph spatiotemporal fusion approach

    Xue Xing et al. “Short -term OD flow predicti on for urban rail transit control: A multi -graph spatiotemporal fusion approach” . In: Information Fusion (2025), p. 102950

  3. [3]

    Deep learning with long short -term memory for time series pred iction

    Li R Hua Y Zhao Z et al. “Deep learning with long short -term memory for time series pred iction” . In: IEEE Communications Magazine 57.6 (2019), pp. 114–119

  4. [4]

    Uniformity and deformation: A benchmark for multi -fish real -time tracking in the farming

    Jinze Huang et al. “Uniformity and deformation: A benchmark for multi -fish real -time tracking in the farming”. In: Expert Systems with Applications 264 (2025), p. 125653

  5. [5]

    A critical review of RNN and LSTM variants in hydrological time series predictions

    Muhammad Waqas and Usa Wannasingha Humphries. “A critical review of RNN and LSTM variants in hydrological time series predictions” . In: MethodsX (2024), p. 102946

  6. [6]

    IoT network traffic classification: a deep learning method with Fourier transform -assisted hyperparameter optimization

    Liu R Xiong Y Dong S et al. “IoT network traffic classification: a deep learning method with Fourier transform -assisted hyperparameter optimization” . In: Frontiers in Physics 11 (2023), p. 1273862

  7. [7]

    Rethinking Fourier Transform from A Basis Functions Perspective for Long- term Time Series Forecasting

    Runze Yang, Longbing Cao, JIE YANG, et al. “Rethinking Fourier Transform from A Basis Functions Perspective for Long- term Time Series Forecasting”. In: Advances in Neural Information Processing Systems 37 (2024), pp. 8515–8540

  8. [8]

    Network Traffic Prediction Based on Short- Time Fourier Transform and Koopa

    Hanwen Cui and Mingju Gong. “Network Traffic Prediction Based on Short- Time Fourier Transform and Koopa”. In: 2024 3rd International Conference on Cloud Computing, Big Data Application and Software Engineering (CBASE). IEEE. 2024, pp. 163–166

Show all 87 references
  1. [9]

    Deep Learning in Carbon Neutrality Forecasting: A Study on the SSA-Attention-BIGRU Network

    Jiwei Ran, Ganchang Zou, and Ying Niu. “Deep Learning in Carbon Neutrality Forecasting: A Study on the SSA-Attention-BIGRU Network”. In: Journal of Organizational and End User Computing (JOEUC) 36.1 (2024), pp. 1–23

  2. [10]

    Short -term traffic flow forecast based on ARIMA -SVM combined model

    Jiaxin Peng, Yongneng Xu, and Menghui Wu. “Short -term traffic flow forecast based on ARIMA -SVM combined model”. In: International Conference on Green Intelligent Transportation System and Safety. Springer. 2021, pp. 287–300

  3. [11]

    An empirical study on personalized product recommendation based on cross -border e-commerce customer data analysis

    Wanwan Li et al . “An empirical study on personalized product recommendation based on cross -border e-commerce customer data analysis” . In: Journal of Organizational and End User Computing (JOEUC) 36.1 (2024), pp. 1–16

  4. [12]

    Robustness verification for knowledge- based logic of risky driving scenes

    Xia Wang et al. “Robustness verification for knowledge- based logic of risky driving scenes”. In: Future of Information and Communication Conference. Springer. 2025, pp. 572–585

  5. [13]

    Intelligent traffic flow prediction using optimized GRU model

    Basharat Hussain et al. “Intelligent traffic flow prediction using optimized GRU model”. In: IEEE Access 9 (2021), pp. 100736–100746

  6. [14]

    Optimizing supply chain management through BO -CNN- LSTM for demand forecasting and inventory management

    Rong Liu and Vina y Vakharia. “Optimizing supply chain management through BO -CNN- LSTM for demand forecasting and inventory management”. In: Journal of Organizational and End User Computing (JOEUC) 36.1 (2024), pp. 1–25

  7. [15]

    Transformer learning for traffic prediction in mobile networks

    Daniel Wass. Transformer learning for traffic prediction in mobile networks. 2021

  8. [16]

    Multi class SVM algorithm with active learning for network traffic classification

    Shi Dong. “Multi class SVM algorithm with active learning for network traffic classification”. In: Expert Systems with Applications 176 (2021), p. 114885

  9. [17]

    The Headaches of Developing a Concussion App for Youth: Balancing Clinical Goals and Technology

    Carol DeMatteo et al. “The Headaches of Developing a Concussion App for Youth: Balancing Clinical Goals and Technology” . In: International Journal of E-Health and Medical Communications (IJEHMC) 15.1 (2024), pp. 1–20

  10. [18]

    An Attention and Wavelet Based Spatial -Temporal Graph Neural Network for Traffi c Flow and Speed Prediction

    Shihao Zhao, Shuli Xing, and Guojun Mao. “An Attention and Wavelet Based Spatial -Temporal Graph Neural Network for Traffi c Flow and Speed Prediction”. In: Mathematics 10.19 (2022), p. 3507

  11. [19]

    Strategic Focus, Tasks, and Paths for Promoting New Productive Forces to Advance Chinese-style Modernization

    Baoping Ren and Ziyue Wang. “Strategic Focus, Tasks, and Paths for Promoting New Productive Forces to Advance Chinese-style Modernization”. In: Journal of Xi’an University of Finance and Economics 37.01 (2024), pp. 3–11

  12. [20]

    Timesnet: Temporal 2d -variation modeling for general time series analysis

    Haixu Wu et al. “Timesnet: Temporal 2d -variation modeling for general time series analysis”. In: arXiv preprint arXiv:2210.02186 (2022)

  13. [21]

    Cross -modal knowledge transfer for 3D point clouds via graph offset prediction

    Huang Zhang et al. “Cross -modal knowledge transfer for 3D point clouds via graph offset prediction” . In: Pattern Recognition 162 (2025), p. 111351

  14. [22]

    Optimization of automated garbage recognition model based on resnet- 50 and weakly supervised cnn for sustainable urban development

    Yan Zhou et al. “Optimization of automated garbage recognition model based on resnet- 50 and weakly supervised cnn for sustainable urban development”. In: Alexandria Engineering Journal 108 (2024), pp. 415–427

  15. [23]

    E-commerce review sentiment analysis and purchase intention prediction based on d eep learning technology

    Xiaoye Ma, Yanyan Li, and Muhammad Asif. “E-commerce review sentiment analysis and purchase intention prediction based on d eep learning technology”. In: Journal of Organizational and End User Computing (JOEUC) 36.1 (2024), pp. 1–29

  16. [24]

    Crop prediction model using machine learning algorithm s

    Topcu A E Elbasi E Zaki C et al. “Crop prediction model using machine learning algorithm s” . In: Applied Sciences 13.16 (2023), p. 9288

  17. [25]

    Self -superv ised learning -based multi -source spectral fusion for fruit qual ity evaluation: A case study in mango fruit ripeness Transactions and Journals 16 prediction

    Liu Zhang et al. “Self -superv ised learning -based multi -source spectral fusion for fruit qual ity evaluation: A case study in mango fruit ripeness Transactions and Journals 16 prediction” . In: Information Fusion 117 (2025), p. 102814

  18. [26]

    A hybrid traffic s peed forecasting approach integrating wavelet transform and motif -based graph convolutional recurre nt neural network

    Na Zhang et al. “A hybrid traffic s peed forecasting approach integrating wavelet transform and motif -based graph convolutional recurre nt neural network”. In: arXiv preprint arXiv:1904.06656 (2019)

  19. [27]

    Rethinking graph neural networks for anomaly detection

    Jianheng Tang et al. “Rethinking graph neural networks for anomaly detection” . In: International conference on machine learning. PMLR. 2022, pp. 21076–21089

  20. [28]

    Detection of Kidney Diseases: Importance of Feature Selection and Classifiers

    Waheeda I Almayyan and Bareeq A AlGhannam. “Detection of Kidney Diseases: Importance of Feature Selection and Classifiers”. In: International Journal of E-Health and Medical Communications (IJEHMC) 15.1 (2024), pp. 1–21

  21. [29]

    Cross -Checking- Based Trademark Image Retrieval for Hot Company Detection

    Hao Wu, Zhiyi Zhang, and Zhilin Zhu. “Cross -Checking- Based Trademark Image Retrieval for Hot Company Detection” . In: Journal of Organizational and End User Computing (JOEUC) 36.1 (2024), pp. 1–12

  22. [30]

    Bilateral Feature Fusion with hexagonal attention for robust saliency detection under uncertain environments

    Habib Khan, Muhammad Talha Usman, and JaKeoung Koo. “Bilateral Feature Fusion with hexagonal attention for robust saliency detection under uncertain environments” . In: Information Fusion (2025), p. 103165

  23. [31]

    Attention enhance d machine instinctive vision with human -inspired saliency detection

    Habib Khan et al. “Attention enhance d machine instinctive vision with human -inspired saliency detection”. In: Image and Vision Computing 152 (2024), p. 105308

  24. [32]

    Analysis and detection against network attacks in the overl apping phenomenon of behavior attribute

    Jiang Xie et al. “ Analysis and detection against network attacks in the overl apping phenomenon of behavior attribute”. In: Computers & Security 121 (2022), p. 102867

  25. [33]

    DDoS attack detection using heuristics clustering algorithm and naıve bayes classification

    Sharmila Bista, Roshan Chitrakar, et al. “DDoS attack detection using heuristics clustering algorithm and naıve bayes classification”. In: Journal of Information Security 9.01 (2017), p. 33

  26. [34]

    Multi -Agent Reinforcement Learning for task allocation in the Internet of Vehicles: Exploring benefits and paving the future

    Inam Ullah et al. “Multi -Agent Reinforcement Learning for task allocation in the Internet of Vehicles: Exploring benefits and paving the future”. In: Swarm and Evolutionary Computation 94 (2025), p. 101878

  27. [35]

    Performance metrics for the comparative analysis of clinical risk prediction models employing machine learning

    Caraballo C Huang C Li S X et al. “Performance metrics for the comparative analysis of clinical risk prediction models employing machine learning”. In: Circulation: Cardiovascular Quality and Outcomes 14.10 (2021), e007526

  28. [36]

    Temporal fusion transformers for interpretable multi -horizon time series forecasting

    Bryan Lim et al. “Temporal fusion transformers for interpretable multi -horizon time series forecasting”. In: International Journal of Forecasting 37.4 (2021), pp. 1748–1764

  29. [37]

    Informer: Beyond efficient transformer for long sequence time-series forecasting

    Peng J Zhou H Zhang S et al. “Informer: Beyond efficient transformer for long sequence time-series forecasting”. In: Proceedings of the AAAI conference on artificial intelligence. Vol. 35. 12. 2021, pp. 11106–11115

  30. [38]

    N-BEATS: Neural basis expansion analysis for interpr etable time series forecasting

    Boris N Oreshkin et al. “N-BEATS: Neural basis expansion analysis for interpr etable time series forecasting”. In : arXiv preprint arXiv :1905.10437 (2019)

  31. [39]

    Autoformer: Decomposition transformers with auto -correlation for lon g-term series forecasting

    Haixu Wu et al. “Autoformer: Decomposition transformers with auto -correlation for lon g-term series forecasting”. In: Advances in neural information processing systems 34 (2021), pp. 22419–22430

  32. [40]

    Probabilistic load forecasting with generative models

    Benjamin Delvoye et al. “ Probabilistic load forecasting with generative models” . In: MatheO (2022)

  33. [41]

    Ubiquitous computation in internet of vehicles for human- centric transport systems

    Inam Ullah et al. “Ubiquitous computation in internet of vehicles for human- centric transport systems”. In: Computers inHuman Behavior 161 (2024), p. 108394

  34. [42]

    Visionary vigilance: Optimized YOLOV8 for fallen person detection with large- scale benchmark dataset

    Habib Khan et al. “Visionary vigilance: Optimized YOLOV8 for fallen person detection with large- scale benchmark dataset”. In: Image and Vision Computing 149 (2024), p. 105195

  35. [43]

    Data-efficient Trajectory Prediction via Coreset Selection

    R. Yang and L. Su, "Data-efficient Trajectory Prediction via Coreset Selection", arXiv preprint arXiv:2409.17385, 2024

  36. [44]

    Electronic Health Records-Based Data-Driven Diabetes Knowledge Unveiling and Risk Prognosis

    H. Pang, L. Zhou, Y. Dong, P. Chen, D. Gu, T. Lyu, and H. Zhang, "Electronic Health Records-Based Data-Driven Diabetes Knowledge Unveiling and Risk Prognosis", IECE Transactions on Intelligent Systematics, 2024, vol. 2, no. 1, pp. 1--13

  37. [45]

    Application of Deep Learning Models Based on EfficientDet and OpenPose in User- Oriented Motion Rehabilitation Robot Control

    M. Sui, L. Jiang, T. Lyu, H. Wang, L. Zhou, P. Chen, and A. Alhosain, "Application of Deep Learning Models Based on EfficientDet and OpenPose in User- Oriented Motion Rehabilitation Robot Control", Journal of Intelligence Technology and Innovation, 2024, vol. 2, no. 3, pp. 47--77

  38. [46]

    Integrating IoT data and reinforcement learning for adaptive macroeconomic policy optimization

    C. Peng, Y. Zhang, and L. Jiang, "Integrating IoT data and reinforcement learning for adaptive macroeconomic policy optimization", Alexandria Engineering Journal, 2025, vol. 119, pp. 222--231

  39. [47]

    EITNet: An IoT- enhanced framework for real-time basketball action recognition

    J. Liu, X. Liu, M. Qu, and T. Lyu, "EITNet: An IoT- enhanced framework for real-time basketball action recognition", Alexandria Engineering Journal, 2025, vol. 110, pp. 567--578

  40. [48]

    GTA-Net: An IoT-integrated 3D human pose estimation system for real-time adolescent sports posture correction

    S. Yuan and L. Zhou, "GTA-Net: An IoT-integrated 3D human pose estimation system for real-time adolescent sports posture correction", Alexandria Engineering Journal, 2025, vol. 112, pp. 585--597

  41. [49]

    The Design of Autonomous UAV Prototypes for Inspecting Tunnel Construction Environment

    Y. Dong, "The Design of Autonomous UAV Prototypes for Inspecting Tunnel Construction Environment", Journal of Intelligence Technology and Innovation, 2024, vol. 2, no. 3, pp. 1--18

  42. [50]

    Risk Analysis in Customer Relationship Management via QRCNN-LSTM and Cross-Attention Mechanism

    Y. Huang, J. Der Leu, B. Lu, and Y. Zhou, "Risk Analysis in Customer Relationship Management via QRCNN-LSTM and Cross-Attention Mechanism", Journal of Organizational and End User Computing (JOEUC), 2024, vol. 36, no. 1, pp. 1--22

  43. [51]

    Deep Analysis of Time Series Data for Smart Grid Startup Strategies: A Transformer-LSTM- PSO Model Approach

    Z. Zhang, "Deep Analysis of Time Series Data for Smart Grid Startup Strategies: A Transformer-LSTM- PSO Model Approach", Journal of Management Science and Operations, 2024, vol. 2, no. 3, pp. 16-- 43

  44. [52]

    Regional BDS satellite clock estimation with triple- frequency ambiguity resolution based on undifferenced observation

    X. Yang, S. Gu, X. Gong, W. Song, Y. Lou, and J. Liu, "Regional BDS satellite clock estimation with triple- frequency ambiguity resolution based on undifferenced observation", GPS Solutions, 2019, vol. 23, pp. 1--11. Transactions and Journals 16

  45. [53]

    Deep Learning-based Anomaly Detection and Log Analysis for Computer Networks

    S. Wang, R. Jiang, Z. Wang, and Y. Zhou, "Deep Learning-based Anomaly Detection and Log Analysis for Computer Networks", Journal of Information and Computing, 2024, vol. 2, no. 2, pp. 34--63

  46. [54]

    PLoc: A New Evaluation Criterion Based on Physical Location for Autonomous Driving Datasets

    R. Yang and Y. Peng, "PLoc: A New Evaluation Criterion Based on Physical Location for Autonomous Driving Datasets", in 2024 12th International Conference on Intelligent Control and Information Processing (ICICIP), 2024, pp. 116--122

  47. [55]

    DPMPC- Planner: A real-time UAV trajectory planning framework for complex static environments with dynamic obstacles

    Z. Xu, D. Deng, Y. Dong, and K. Shimada, "DPMPC- Planner: A real-time UAV trajectory planning framework for complex static environments with dynamic obstacles", in 2022 International Conference on Robotics and Automation (ICRA), 2022, pp. 250-- 256

  48. [56]

    Anomaly Detection and Risk Early Warning System for Financial Time Series Based on the WaveLST- Trans Model

    T. Su, R. Li, B. Liu, X. Liang, X. Yang, and Y. Zhou, "Anomaly Detection and Risk Early Warning System for Financial Time Series Based on the WaveLST- Trans Model", 2025

  49. [57]

    Short and Long-Term Renewable Electricity Demand Forecasting Based on CNN-Bi-GRU Model

    S. Zhao, Z. Xu, Z. Zhu, X. Liang, Z. Zhang, and R. Jiang, "Short and Long-Term Renewable Electricity Demand Forecasting Based on CNN-Bi-GRU Model", IECE Transactions on Emerging Topics in Artificial Intelligence, 2025, vol. 2, no. 1, pp. 1--15

  50. [58]

    Intelligent design and optimization of exercise equipment based on fusion algorithm of YOLOv5- ResNet 50

    L. Wang, W. Ji, G. Wang, Y. Feng, and M. Du, "Intelligent design and optimization of exercise equipment based on fusion algorithm of YOLOv5- ResNet 50", Alexandria Engineering Journal, 2024, vol. 104, pp. 710--722

  51. [59]

    DR-CircuitGNN: Training Acceleration of Heterogeneous Circuit Graph Neural Network on GPUs

    Y. Luo, S. Li, J. Tao, K. G. Thorat, X. Xie, H. Peng, N. Xu, C. Ding, and S. Huang, "DR-CircuitGNN: Training Acceleration of Heterogeneous Circuit Graph Neural Network on GPUs", 2025

  52. [60]

    Multi-GNSS satellite clock estimation constrained with oscillator noise model in the existence of data discontinuity

    C. Shi, S. Guo, S. Gu, X. Yang, X. Gong, Z. Deng, M. Ge, and H. Schuh, "Multi-GNSS satellite clock estimation constrained with oscillator noise model in the existence of data discontinuity", Journal of Geodesy, 2019, vol. 93, pp. 515--528

  53. [61]

    TRIZ Method for Urban Building Energy Optimization: GWO-SARIMA-LSTM Forecasting model

    S. Zheng, S. Liu, Z. Zhang, D. Gu, C. Xia, H. Pang, and E. M. Ampaw, "TRIZ Method for Urban Building Energy Optimization: GWO-SARIMA-LSTM Forecasting model", Journal of Intelligence Technology and Innovation, 2024, vol. 2, no. 3, pp. 78--103

  54. [62]

    Enhancing human pose estimation in sports training: Integrating spatiotemporal transformer for improved accuracy and real-time performance

    X. Xi, C. Zhang, W. Jia, and R. Jiang, "Enhancing human pose estimation in sports training: Integrating spatiotemporal transformer for improved accuracy and real-time performance", Alexandria Engineering Journal, 2024, vol. 109, pp. 144--156

  55. [63]

    MiM-UNet: An efficient building image segmentation network integrating state space models

    D. Liu, Z. Wang, and A. Liang, "MiM-UNet: An efficient building image segmentation network integrating state space models", Alexandria Engineering Journal, 2025, vol. 120, pp. 648--656

  56. [64]

    Image anomaly detection and prediction scheme based on SSA optimized ResNet50-BiGRU model

    Q. Wan, Z. Zhang, L. Jiang, Z. Wang, and Y. Zhou, "Image anomaly detection and prediction scheme based on SSA optimized ResNet50-BiGRU model", arXiv preprint arXiv:2406.13987, 2024

  57. [65]

    Impact of Short- Duration Aerobic Exercise Intensity on Executive Function and Sleep

    Y. Peng, G. Zhang, and H. Pang, "Impact of Short- Duration Aerobic Exercise Intensity on Executive Function and Sleep", arXiv preprint arXiv:2503.09077, 2025

  58. [66]

    Application of Deep Learning for Automatic Identification of Hazardous Materials and Urban Safety Supervision

    T. Yan, J. Wu, M. Kumar, and Y. Zhou, "Application of Deep Learning for Automatic Identification of Hazardous Materials and Urban Safety Supervision", Journal of Organizational and End User Computing (JOEUC), 2024, vol. 36, no. 1, pp. 1--20

  59. [67]

    Intelligent logistics management robot path planning algorithm integrating transformer and GCN network

    H. Luo, J. Wei, S. Zhao, A. Liang, Z. Xu, and R. Jiang, "Intelligent logistics management robot path planning algorithm integrating transformer and GCN network", IECE Transactions on Internet of Things, 2024, vol. 2, pp. 95-112

  60. [68]

    The effect of intelligent monitoring of physical exercise on executive function in children with ADHD

    L. Lin, N. Li, and S. Zhao, "The effect of intelligent monitoring of physical exercise on executive function in children with ADHD", Alexandria Engineering Journal, 2025, vol. 122, pp. 355--363

  61. [69]

    Real-time monitoring of lower limb movement resistance based on deep learning

    Y. Liu, T. Lyu, and . others, "Real-time monitoring of lower limb movement resistance based on deep learning", Alexandria Engineering Journal, 2025, vol. 111, pp. 136--147

  62. [70]

    LTPNet Integration of Deep Learning and Environmental Decision Support Systems for Renewable Energy Demand Forecasting

    T. Li, M. Zhang, and Y. Zhou, "LTPNet Integration of Deep Learning and Environmental Decision Support Systems for Renewable Energy Demand Forecasting", arXiv preprint arXiv:2410.15286, 2024

  63. [71]

    Recording brain activity while listening to music using wearable EEG devices combined with Bidirectional Long Short-Term Memory Networks

    J. Wang, Z. Wang, and G. Liu, "Recording brain activity while listening to music using wearable EEG devices combined with Bidirectional Long Short-Term Memory Networks", Alexandria Engineering Journal, 2024, vol. 109, pp. 1--10

  64. [72]

    Automatic News Generation and Fact-Checking System Based on Language Processing

    X. Peng, Q. Xu, Z. Feng, H. Zhao, L. Tan, Y. Zhou, Z. Zhang, C. Gong, and Y. Zheng, "Automatic News Generation and Fact-Checking System Based on Language Processing", Journal of Industrial Engineering and Applied Science, 2024, vol. 2, no. 3, pp. 1--11

  65. [73]

    Cross-border Commodity Pricing Strategy Optimization via Mixed Neural Network for Time Series Analysis

    L. Wang, Y. Hu, and Y. Zhou, "Cross-border Commodity Pricing Strategy Optimization via Mixed Neural Network for Time Series Analysis", arXiv preprint arXiv:2408.12115, 2024

  66. [74]

    Optimized CNNs for Rapid 3D Point Cloud Object Recognition

    T. Lyu, D. Gu, P. Chen, Y. Jiang, Z. Zhang, H. Pang, L. Zhou, and Y. Dong, "Optimized CNNs for Rapid 3D Point Cloud Object Recognition", IECE Transactions on Internet of Things, 2024, vol. 2, no. 4, pp. 83--94

  67. [75]

    Theoretical Analysis of Meta Reinforcement Learning: Generalization Bounds and Convergence Guarantees

    C. Wang, M. Sui, D. Sun, Z. Zhang, and Y. Zhou, "Theoretical Analysis of Meta Reinforcement Learning: Generalization Bounds and Convergence Guarantees", 2024, pp. 153–159, doi: 10.1145/3677779.3677804

  68. [76]

    Unsupervised Search for Ethnic Minorities' Medical Segmentation Training Set

    T. Gedeon, Y. Chen, Y. Yao, R. Yang, M. Z. Hossain, and A. Gupta, "Unsupervised Search for Ethnic Minorities' Medical Segmentation Training Set", arXiv preprint, 2025

  69. [77]

    Enhancing Visual Question Answering through Ranking-Based Hybrid Training and Multimodal Fusion

    P. Chen, Z. Zhang, Y. Dong, L. Zhou, and H. Wang, "Enhancing Visual Question Answering through Ranking-Based Hybrid Training and Multimodal Fusion", Journal of Intelligence Technology and Innovation, 2024, vol. 2, no. 3, pp. 19--46

  70. [78]

    Real-time monitoring and analysis of track and field athletes based on edge computing and deep reinforcement learning algorithm

    X. Tang, B. Long, and L. Zhou, "Real-time monitoring and analysis of track and field athletes based on edge computing and deep reinforcement learning algorithm", Alexandria Engineering Journal, 2025, vol. 114, pp. 136--146

  71. [79]

    IoT-based 3D pose estimation and motion optimization for athletes: Application of C3D and OpenPose

    F. Ren, C. Ren, and T. Lyu, "IoT-based 3D pose estimation and motion optimization for athletes: Application of C3D and OpenPose", Alexandria Engineering Journal, 2025, vol. 115, pp. 210--221

  72. [80]

    Optimization of automated Transactions and Journals 16 garbage recognition model based on ResNet-50 and weakly supervised CNN for sustainable urban development

    Y. Zhou, Z. Wang, S. Zheng, L. Zhou, L. Dai, H. Luo, Z. Zhang, and M. Sui, "Optimization of automated Transactions and Journals 16 garbage recognition model based on ResNet-50 and weakly supervised CNN for sustainable urban development", Alexandria Engineering Journal, 2024, v...

  73. [81]

    RTop-K: Ultra- Fast Row-Wise Top-K Selection for Neural Network Acceleration on GPUs

    X. Xie, Y. Luo, H. Peng, and C. Ding, "RTop-K: Ultra- Fast Row-Wise Top-K Selection for Neural Network Acceleration on GPUs", in The Thirteenth International Conference on Learning Representations

  74. [82]

    CPLOYO: A pulmonary nodule detection model with multi-scale feature fusion and nonlinear feature learning

    M. Wang, Z. Yang, R. Zhao, and Y. Jiang, "CPLOYO: A pulmonary nodule detection model with multi-scale feature fusion and nonlinear feature learning", Alexandria Engineering Journal, 2025, vol. 122, pp. 578--587

  75. [83]

    Research on empirical correction models of GPS Block IIF and BDS satellite inter-frequency clock bias

    X. Gong, S. Gu, Y. Lou, F. Zheng, X. Yang, Z. Wang, and J. Liu, "Research on empirical correction models of GPS Block IIF and BDS satellite inter-frequency clock bias", Journal of Geodesy, 2020, vol. 94, pp. 1--11

  76. [84]

    Federated learning application on depression treatment robots (DTbot)

    Y. Liu and R. Yang, "Federated learning application on depression treatment robots (DTbot)", in 2021 IEEE 13th International Conference on Computer Research and Development (ICCRD), 2021, pp. 121--124

  77. [85]

    Construction and optimization of health behavior prediction model for the older adult in smart older adult care

    Q. Guo and P. Chen, "Construction and optimization of health behavior prediction model for the older adult in smart older adult care", Frontiers in Public Health, 2024, vol. 12, pp. 1486930

  78. [86]

    Rumor Detection of Sina Weibo Based on MCF Algorithm

    Y. Liu and R. Yang, "Rumor Detection of Sina Weibo Based on MCF Algorithm", in 2020 International Conference on Computing and Data Science (CDS), 2020, pp. 411--414

  79. [87]

    DSEM-NeRF: Multimodal feature fusion and global-local attention for enhanced 3D scene reconstruction

    D. Liu, Z. Wang, and P. Chen, "DSEM-NeRF: Multimodal feature fusion and global-local attention for enhanced 3D scene reconstruction", Information Fusion, 2024, pp. 102752

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.