Pith. sign in

REVIEW 4 major objections 4 minor 1 cited by

Artificial Intelligence-Based Multiscale Temporal Modeling for Anomaly Detection in Cloud Services

T0 review · 4 major / 4 minor · reviewed 2026-08-05 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read This paper claims that a Transformer with multiscale downsampling paths and attention-weighted fusion outperforms six baseline models at detecting anomalies in cloud monitoring time series.

desk verdict Plausible multiscale Transformer recombination, but every reported number rests on anomaly labels that the paper never shows exist in Alibaba Cluster Trace 2018. read the letter →

arxiv 2508.14503 v2 pith:RFMUNRTK submitted 2025-08-20 cs.LG

Lian Lian , Yilin Li , Song Han , Renzi Meng , Sibo Wang , Ming Wang This is my paper · ORCID
classification cs.LG
keywords anomalydetectioncloudservicesTransformermultiscalefeaturefusionself-attentiontimeseriesAlibabaClusterTrace2018robustnessanalysis
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

This paper argues that a Transformer can be made substantially better at detecting anomalies in cloud monitoring data by processing the time series at several downsampled temporal scales and letting an attention mechanism weight how much each scale contributes. On the Alibaba Cluster Trace 2018 dataset, the proposed model reports an F1 score of 0.894 and an AUC of 0.941, above a plain Transformer, a Transformer+CNN hybrid, LSTM, GRU, ResNet, and VGG. The design targets the practical need to catch both short-lived spikes and slow-moving trends in CPU, memory, and scheduling signals. The paper also shows how learning rate, optimizer, noise, and anomaly proportion shift precision and recall, which matters for deploying such a detector in real cloud operations.

What carries the argument

The central object is an improved Transformer encoder surrounded by a multiscale fusion path. The input time series is downsampled at several scales; each scale is encoded by its own Transformer module; a scale-alignment step remaps all representations to a common length and dimension; and an attention-weighted fusion module learns a weight per scale and sums the aligned features into a final representation. This lets the model attend to both fine-grained fluctuations and coarse-grained trends at once, and the learned weights are what adapt the fusion to each input.

What would settle it

Inspect the public Alibaba Cluster Trace 2018 release: it contains job, task, and container workload records but no per-instance labels for the anomaly events the paper lists (service interruptions, resource surges, latency drifts). Since the paper's label-refinement step is unspecified and no labeled dataset is released, there is no way to compute the reported precision, recall, AUC, or F1 from the public data; this would settle whether the central comparison is reproducible.

Watch

Extended reading notes

Core claim

The paper's central claim is that fusing multiple temporal resolutions inside a Transformer improves anomaly detection in cloud services. It constructs several downsampled views of the input sequence, encodes each view with its own Transformer module, aligns the encoded features back to a common shape, and fuses them with learned attention weights. On Alibaba Cluster Trace 2018 it reports the best metrics among the tested methods: precision 0.902, recall 0.887, AUC 0.941, and F1 0.894. The attention-weighted fusion, rather than any single scale, is what the paper credits for balancing local detail with long-term context.

Load-bearing premise

The load-bearing premise is that the 'label refinement' step applied to the Alibaba Cluster Trace 2018 produces valid anomaly labels; the public trace does not contain per-event labels for service interruptions, resource surges, or latency drifts, and the paper does not state how those labels were created.

Editorial extensions

If this is right

  • On the Alibaba Cluster Trace 2018 benchmark, the proposed multiscale Transformer reports the highest precision (0.902), recall (0.887), AUC (0.941), and F1 (0.894) among the compared methods.
  • Attention-weighted fusion lets the model adjust each scale's contribution per input, which is why the paper says it can track both short-term fluctuations and long-term trends.
  • The sensitivity analysis identifies AdamW with learning rate 1e-4 as the best configuration; for deployment, these settings matter more than the architecture alone.
  • Under injected noise, recall and F1 decline fastest, so the claimed stability has limits; and under rare anomalies (1-5%), precision stays high while recall drops, with F1 peaking near a 10% anomaly ratio.

Reading between the lines

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

  • Because the architecture consumes arbitrary multivariate time series, the same multiscale Transformer could be carried over to other telemetry domains, such as network or storage monitoring, as long as per-instance anomaly labels are available.
  • The paper's noise-injection results imply that denoising or perturbation-aware training is a natural next step; whether it preserves the reported F1 is a direct experiment.
  • The conclusion's concern about quadratic attention cost suggests testing sparse or linear attention variants; their effect on the precision/recall trade-off is an open question.
  • A rerun on a benchmark with public anomaly labels, such as Exathlon—which the paper names as future work—would test whether the reported gains generalize beyond the Alibaba trace.
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 / 4 minor

Summary. The paper proposes a multiscale Transformer architecture with attention-weighted fusion for anomaly detection in cloud services. The method applies a standard Transformer encoder to multidimensional monitoring sequences, constructs multiple downsampled scales, aligns them, and fuses the scale-specific representations with learned attention weights. The authors evaluate on Alibaba Cluster Trace 2018 and report that the proposed method outperforms ResNet, VGG, LSTM, GRU, Transformer, and Transformer+CNN, achieving precision 0.902, recall 0.887, AUC 0.941, and F1 0.894. Additional experiments study sensitivity to learning rate, optimizer, activation function, noise, anomaly ratio, and training loss.

Significance. If the empirical claim were verified, the main contribution would be a straightforward but plausible combination of multiscale feature extraction and Transformer-based temporal modeling, showing gains on a widely used cloud trace. The paper explicitly acknowledges computational and generalization limitations in Section VI, and the sensitivity analysis is a useful step. However, the paper does not ship code, does not release the constructed dataset or labels, omits central architecture details, and reports no uncertainty or significance testing. The positive aspects--transparent limitation statements and a systematic sensitivity view--do not offset the absence of verifiable evidence for the headline claim.

major comments (4)
  1. [Section IV.A] The entire evaluation rests on anomaly labels whose origin is not documented. The paper states that Alibaba Cluster Trace 2018 includes 'annotated anomaly events such as service interruptions, resource surges, and latency drifts' and that preprocessing includes 'label refinement,' but it never specifies the labeling rule, the labeler, the label distribution, or how false labels were controlled, and no label file or labeling script is released. The public Alibaba Cluster Trace 2018 is a workload trace with job/task/container events; it does not, to this reviewer's knowledge, contain per-instance anomaly annotations of the kind described. If the labels were produced by an undocumented heuristic or are otherwise misaligned, every precision, recall, AUC, and F1 value in Tables 1-3 is uninterpretable, and the claimed superiority over the baselines is unsupported. This concern cannot be resolv
  2. [Section III.A-B] The method cannot be reproduced from the text. The number of scales S, the downsampling function φ, the hidden dimension d, the number of Transformer layers and heads, the feed-forward width, and the alignment function f_s are all unspecified. Several equations are corrupted by typesetting errors (e.g., the definitions of X, H^(0), and the attention expressions), making it impossible to infer the exact computation. For an empirical paper whose central claim is 'ours outperforms six baselines,' this omission is load-bearing: a reviewer cannot tell whether the reported numbers correspond to the described architecture or to a different configuration.
  3. [Section V.A, Table 1] The comparative claim is reported without any measure of uncertainty. The text says models were 'trained multiple times independently under the same data split, hyperparameter settings, and random seed,' but Table 1 reports single-point metrics with no variance, confidence intervals, or significance tests. Moreover, 'same random seed' does not yield independent replicates. The differences between the proposed method and the strongest baseline (F1 0.894 vs. 0.871, AUC 0.941 vs. 0.923) may be real, but as reported they could be within run-to-run variability. The claim that the method 'confirms its robustness, adaptability, and suitability' is therefore not statistically supported.
  4. [Section V.B and Section IV.B] The experimental protocol appears to select hyperparameters on the same benchmark that is later used for the headline comparison. The paper reports the best learning rate (1e-4) and optimizer (AdamW) from Tables 2 and 3, and these values are then used for the 'Ours' row in Table 1, but no held-out validation split or nested selection procedure is described. This creates a selection effect: the reported performance is the best over the tested configurations. In addition, Figure 6 varies the 'anomaly ratio' without explaining how the ratio is changed, which is especially problematic given the unverified labels. The authors should describe the selection protocol and, ideally, report results on a truly held-out test set.
minor comments (4)
  1. [General typesetting] Many equations are garbled or unreadable, including the input representation, positional encoding, attention formula, and the multiscale attention-weight expression in Sections III.A and III.B. The paper needs a careful proofread of all mathematical notation.
  2. [Section IV.A] The description of Alibaba Cluster Trace 2018 overstates what the public trace contains. The trace provides job/task/container events and resource utilization, not the 'service interruptions, resource surges, and latency drifts' annotations described. The authors should cite the exact release and describe any additional annotation step transparently.
  3. [Section V.C, Figure 3] The activation-function sensitivity result is presented only as a figure with no numeric values, no axis labels, and no description of how activation was varied across the multiscale Transformer. This makes the claim about GELU difficult to verify.
  4. [Conclusion] The conclusion says 'Without relying on additional label information, the model improves its responsiveness to anomalous regions,' which is inconsistent with the fully supervised training setup implied by Tables 1-3. Clarify whether the model is trained with anomaly labels or not.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity; the empirical claim is compared against external baselines and no fitted quantity is renamed as a prediction.

full rationale

The paper contains no theoretical derivation chain whose output is equivalent to its inputs by construction. The method (Transformer backbone, multiscale downsampling, attention-weighted fusion) is an architectural recombination, not a re-derivation of the benchmark numbers. The central empirical claim is a comparative evaluation against ResNet, VGG, LSTM, GRU, Transformer, and Transformer+CNN on a public trace, so it is externally falsifiable rather than self-referential. Hyperparameter sensitivity experiments (learning rate, optimizer, anomaly ratio, noise) select settings that perform best on the same benchmark; this is a conventional selection effect and is not the pattern of fitting a parameter to a subset and then 'predicting' a closely related quantity. No uniqueness theorem is imported, no ansatz is smuggled in via citation, and the related-work citations are not load-bearing for the reported numbers. The main weakness is evidential rather than circular: Section IV.A asserts that Alibaba Cluster Trace 2018 contains 'annotated anomaly events' and that preprocessing includes 'label refinement,' but the labeling rule is never specified and no label file is released. That threatens interpretability of every metric, but it is a missing-evidence problem, not a circularity problem. A low score is therefore appropriate.

Assumptions & free parameters 6 free parameters · 3 assumptions · 0 invented entities

The load-bearing assumptions are empirical: anomaly-label validity, multiscale information preservation, and distribution transfer. The paper supplies no extra theoretical entities. Its architectural choices, including scales, window size, optimizer, and learning rate, are free parameters, and several of them are not even reported.

free parameters (6)
  • learning_rate = 1e-4
    Selected as best in Table 2 sensitivity sweep; final results use it.
  • optimizer = AdamW
    Selected as best in Table 3 optimizer sweep; final results use it.
  • window_size = 60
    Fixed in Section IV.B without sensitivity analysis.
  • stride = 1
    Fixed in Section IV.B without sensitivity analysis.
  • number_of_scales_S_and_downsampling_function = not reported
    The multiscale construction in Section III.B depends on S and the scaling function phi, which are never specified.
  • transformer_depth_width_heads = not reported
    Encoder layers, hidden dimension, and head count are not stated; these determine model capacity and are effectively chosen by hand.
assumptions (3)
  • domain assumption Alibaba Cluster Trace 2018 contains valid anomaly labels after preprocessing.
    Section IV.A asserts annotated anomaly events and label refinement, but the public trace does not carry such labels and no labeling source is cited.
  • domain assumption Downsampled and aligned multiscale representations preserve sufficient temporal information for fusion.
    Section III.B introduces parallel scales and scale alignment without an ablation or theoretical guarantee.
  • domain assumption Supervised training on windows with consistent anomaly proportions transfers to deployment conditions.
    Section IV.B fixes train/test anomaly proportion; Section VI acknowledges distribution shift may degrade performance.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Artificial Intelligence-Based Multiscale Temporal Modeling for Anomaly Detection in Cloud Services." pith.science (2026). https://pith.science/paper/RFMUNRTK

@misc{pith2026250814503,
  author       = {Pith},
  title        = {Pith review of: Artificial Intelligence-Based Multiscale Temporal Modeling for Anomaly Detection in Cloud Services},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/RFMUNRTK}},
  note         = {Machine review of arXiv:2508.14503}
}
read the original abstract

This study proposes an anomaly detection method based on the Transformer architecture with integrated multiscale feature perception, aiming to address the limitations of temporal modeling and scale-aware feature representation in cloud service environments. The method first employs an improved Transformer module to perform temporal modeling on high-dimensional monitoring data, using a self-attention mechanism to capture long-range dependencies and contextual semantics. Then, a multiscale feature construction path is introduced to extract temporal features at different granularities through downsampling and parallel encoding. An attention-weighted fusion module is designed to dynamically adjust the contribution of each scale to the final decision, enhancing the model's robustness in anomaly pattern modeling. In the input modeling stage, standardized multidimensional time series are constructed, covering core signals such as CPU utilization, memory usage, and task scheduling states, while positional encoding is used to strengthen the model's temporal awareness. A systematic experimental setup is designed to evaluate performance, including comparative experiments and hyperparameter sensitivity analysis, focusing on the impact of optimizers, learning rates, anomaly ratios, and noise levels. Experimental results show that the proposed method outperforms mainstream baseline models in key metrics, including precision, recall, AUC, and F1-score, and maintains strong stability and detection performance under various perturbation conditions, demonstrating its superior capability in complex cloud environments.

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Topology-Aware Graph Reinforcement Learning for Dynamic Routing in Cloud Networks

    cs.LG 2025-09 reject novelty 3.0 of 10

    A graph RL router with structure-aware state encoding and policy-driven edge rewiring reports improved throughput and latency on GEANT, based on single-run table comparisons.

Reference graph

Works this paper leans on

55 extracted references · 43 canonical work pages · cited by 1 Pith paper

  1. [1]

    Decomposition-based multi-scale transformer framework for time series anomaly detection,

    W. Zhang and C. Luo, "Decomposition-based multi-scale transformer framework for time series anomaly detection," Neural Networks, vol. 187, 107399, 2025

    work page 2025

  2. [2]

    Multi-scale transformers with adaptive pathways for time series forecasting,

    P. Chen, Y. Zhang, Y. Cheng, et al., "Multi-scale transformers with adaptive pathways for time series forecasting," Proceedings of the International Conference on Learning Representations, 2024

    work page 2024

  3. [3]

    Graph-Based Representation Learning for Identifying Fraud in Transaction Networks,

    X. Guo, Y. Wu, W. Xu, Z. Liu, X. Du and T. Zhou, "Graph-Based Representation Learning for Identifying Fraud in Transaction Networks," 2025

    work page 2025

  4. [4]

    Reinforcement Learning in Finance: QTRAN for Portfolio Optimization,

    Z. Xu, Q. Bao, Y. Wang, H. Feng, J. Du and Q. Sha, "Reinforcement Learning in Finance: QTRAN for Portfolio Optimization," Journal of Computer Technology and Software, vol. 4, no. 3, 2025

    work page 2025

  5. [5]

    A Deep Learning Approach to Anomaly Detection in High-Frequency Trading Data,

    Q. Bao, J. Wang, H. Gong, Y. Zhang, X. Guo and H. Feng, "A Deep Learning Approach to Anomaly Detection in High-Frequency Trading Data," Proceedings of the 2025 4th International Symposium on Computer Applications and Information Technology (ISCAIT), pp. 287- 291, 2025

    2025

  6. [6]

    Time-Series Nested Reinforcement Learning for Dynamic Risk Control in Nonlinear Financial Markets,

    Y. Yao, "Time-Series Nested Reinforcement Learning for Dynamic Risk Control in Nonlinear Financial Markets," Transactions on Computational and Scientific Methods, vol. 5, no. 1, 2025

    work page 2025

  7. [7]

    Enhancing Systemic Risk Forecasting with Deep Attention Models in Financial Time Series,

    Q. R. Xu, W. Xu, X. Su, K. Ma, W. Sun and Y. Qin, "Enhancing Systemic Risk Forecasting with Deep Attention Models in Financial Time Series," 2025

    work page 2025

  8. [8]

    Transformer-Based Risk Monitoring for Anti-Money Laundering with Transaction Graph Integration,

    Y. Wu, Y. Qin, X. Su and Y. Lin, "Transformer-Based Risk Monitoring for Anti-Money Laundering with Transaction Graph Integration," 2025

    work page 2025

Show all 55 references
  1. [9]

    Stock Prediction with Improved Feedforward Neural Networks and Multimodal Fusion,

    Y. Wang, "Stock Prediction with Improved Feedforward Neural Networks and Multimodal Fusion," Journal of Computer Technology and Software, vol. 4, no. 1, 2025

    2025

  2. [10]

    Market Turbulence Prediction and Risk Control with Improved A3C Reinforcement Learning,

    J. Liu, X. Gu, H. Feng, Z. Yang, Q. Bao and Z. Xu, "Market Turbulence Prediction and Risk Control with Improved A3C Reinforcement Learning," Proceedings of the 2025 8th International Conference on Advanced Algorithms and Control Engineering (ICAACE), pp. 2634- 2638, 2025

    2025

  3. [11]

    Opprentice: Towards practical and automatic anomaly detection through machine learning

    D. Liu, Y. Zhao, H. Xu, Y. Sun, D. Pei, J. Luo, X. Jing and M. Feng, "Opprentice: Towards practical and automatic anomaly detection through machine learning", Proceedings of the 2015 ACM Internet Measurement Conference (IMC), pp. 211-224

    2015

  4. [12]

    Automatic anomaly detection in the cloud via statistical learning

    O. Vallis, J. Hochenbaum and A. Kejariwal, "Automatic anomaly detection in the cloud via statistical learning", arXiv:1704.07706, 2017

    2017 arXiv

  5. [13]

    Evaluating real-time anomaly detection algorithms—The Numenta Anomaly Benchmark

    A. Lavin and S. Ahmad, "Evaluating real-time anomaly detection algorithms—The Numenta Anomaly Benchmark", Proceedings of the 2015 IEEE International Conference on Machine Learning and Applications (ICMLA), 2015

    2015

  6. [14]

    EEG Anomaly Detection Using Temporal Graph Attention for Clinical Applications,

    X. Zhang and Q. Wang, "EEG Anomaly Detection Using Temporal Graph Attention for Clinical Applications," Journal of Computer Technology and Software, vol. 4, no. 7, 2025

    2025

  7. [15]

    Multi-Scale Feature Integration and Spatial Attention for Accurate Lesion Segmentation,

    Y. Wu, Y. Lin, T. Xu, X. Meng, H. Liu and T. Kang, "Multi-Scale Feature Integration and Spatial Attention for Accurate Lesion Segmentation," 2025

    2025

  8. [16]

    Clinical NLP with Attention-Based Deep Learning for Multi-Disease Prediction,

    T. Xu, X. Deng, X. Meng, H. Yang and Y. Wu, "Clinical NLP with Attention-Based Deep Learning for Multi-Disease Prediction," arXiv preprint arXiv:2507.01437, 2025

    2025 arXiv

  9. [17]

    Domain-Adaptive Organ Segmentation through SegFormer Architecture in Clinical Imaging,

    X. Zhang and X. Wang, "Domain-Adaptive Organ Segmentation through SegFormer Architecture in Clinical Imaging," , 2025

    2025

  10. [18]

    Joint Modeling of Medical Images and Clinical Text for Early Diabetes Risk Detection,

    Y. Zi and X. Deng, "Joint Modeling of Medical Images and Clinical Text for Early Diabetes Risk Detection," Journal of Computer Technology and Software, vol. 4, no. 7, 2025

    2025

  11. [19]

    A novel multi-scale network intrusion detection model with transformer,

    C. Xi, H. Wang and X. Wang, "A novel multi-scale network intrusion detection model with transformer," Scientific Reports, vol. 14, no. 1, 23239, 2024

    2024

  12. [20]

    Multirc: Joint learning for time series anomaly prediction and detection with multi-scale reconstructive contrast,

    S. Hu, K. Zhao, X. Qiu, et al., "Multirc: Joint learning for time series anomaly prediction and detection with multi-scale reconstructive contrast," arXiv preprint arXiv:2410.15997, 2024

    2024 arXiv

  13. [21]

    A Hybrid Recommendation Approach Integrating Matrix Decomposition and Deep Neural Networks for Enhanced Accuracy and Generalization,

    R. Wang, Y. Luo, X. Li, Z. Zhang, J. Hu and W. Liu, "A Hybrid Recommendation Approach Integrating Matrix Decomposition and Deep Neural Networks for Enhanced Accuracy and Generalization," Proceedings of the 2025 5th International Conference on Neural Networks, Information and C...

    2025

  14. [22]

    Modeling Multi- Hop Semantic Paths for Recommendation in Heterogeneous Information Networks,

    H. Zheng, Y. Xing, L. Zhu, X. Han, J. Du and W. Cui, "Modeling Multi- Hop Semantic Paths for Recommendation in Heterogeneous Information Networks," arXiv preprint arXiv:2505.05989, 2025

    2025 arXiv

  15. [23]

    Sequential Recommendation via Time- Aware and Multi-Channel Convolutional User Modeling,

    Y. Xing, Y. Wang and L. Zhu, "Sequential Recommendation via Time- Aware and Multi-Channel Convolutional User Modeling," Transactions on Computational and Scientific Methods, vol. 5, no. 5, 2025

    2025

  16. [24]

    EdgeConvFormer: Dynamic Graph CNN and Transformer based Anomaly Detection in Multivariate Time Series,

    J. Liu, Q. Li, S. An, et al., "EdgeConvFormer: Dynamic Graph CNN and Transformer based Anomaly Detection in Multivariate Time Series," arXiv preprint arXiv:2312.01729, 2023

    2023 arXiv

  17. [25]

    Collaborative Distillation Strategies for Parameter-Efficient Language Model Deployment,

    X. Meng, Y. Wu, Y. Tian, X. Hu, T. Kang and J. Du, "Collaborative Distillation Strategies for Parameter-Efficient Language Model Deployment," arXiv preprint arXiv:2507.15198, 2025

    2025 arXiv

  18. [26]

    Revisiting LoRA: A Smarter Low-Rank Approach for Efficient Model Adaptation,

    Y. Peng, Y. Wang, Z. Fang, L. Zhu, Y. Deng and Y. Duan, "Revisiting LoRA: A Smarter Low-Rank Approach for Efficient Model Adaptation," Proceedings of the 2025 5th International Conference on Artificial Intelligence and Industrial Technology Applications (AIITA), pp. 1248- 1252, 2025

    2025

  19. [27]

    Integrating Knowledge Graph Reasoning with Pretrained Language Models for Structured Anomaly Detection,

    X. Liu, Y. Qin, Q. Xu, Z. Liu, X. Guo and W. Xu, "Integrating Knowledge Graph Reasoning with Pretrained Language Models for Structured Anomaly Detection," 2025

    2025

  20. [28]

    Structured Path Guidance for Logical Coherence in Large Language Model Generation,

    X. Quan, "Structured Path Guidance for Logical Coherence in Large Language Model Generation," Journal of Computer Technology and Software, vol. 3, no. 3, 2024

    2024

  21. [29]

    Anomaly detection and analysis for clustered cloud computing reliability,

    A. Samir and C. Pahl, "Anomaly detection and analysis for clustered cloud computing reliability," CLOUD COMPUTING, 2019, 2019: 120

    2019

  22. [30]

    Anomaly detection in microservice environments via conditional multiscale GANs and adaptive temporal autoencoders,

    Y. Ma, "Anomaly detection in microservice environments via conditional multiscale GANs and adaptive temporal autoencoders," Transactions on Computational and Scientific Methods, vol. 4, no. 10, 2024

    2024

  23. [31]

    Self-Attention-Based Modeling of Multi-Source Metrics for Performance Trend Prediction in Cloud Systems,

    H. Xin and R. Pan, "Self-Attention-Based Modeling of Multi-Source Metrics for Performance Trend Prediction in Cloud Systems," Journal of Computer Technology and Software, vol. 4, no. 4, 2025

    2025

  24. [32]

    Graph Neural Recognition of Malicious User Patterns in Cloud Systems via Attention Optimization,

    D. Gao, "Graph Neural Recognition of Malicious User Patterns in Cloud Systems via Attention Optimization," Transactions on Computational and Scientific Methods, vol. 4, no. 12, 2024

    2024

  25. [33]

    Deep Learning for Root Cause Detection in Distributed Systems with Structural Encoding and Multi-modal Attention,

    Y. Ren, "Deep Learning for Root Cause Detection in Distributed Systems with Structural Encoding and Multi-modal Attention," Journal of Computer Technology and Software, vol. 3, no. 5, 2024

    2024

  26. [34]

    PatchTrAD: A Patch-Based Transformer focusing on Patch-Wise Reconstruction Error for Time Series Anomaly Detection,

    S. M. Vilhes, G. Gasso and M. Z. Alaya, "PatchTrAD: A Patch-Based Transformer focusing on Patch-Wise Reconstruction Error for Time Series Anomaly Detection," arXiv preprint arXiv:2504.08827, 2025

    2025 arXiv

  27. [35]

    A Deep Learning Framework for Sequence Mining with Bidirectional LSTM and Multi-Scale Attention,

    T. Yang, Y. Cheng, Y. Ren, Y. Lou, M. Wei and H. Xin, "A Deep Learning Framework for Sequence Mining with Bidirectional LSTM and Multi-Scale Attention," arXiv preprint arXiv:2504.15223, 2025

    2025 arXiv

  28. [36]

    Multi-scale feature reconstruction network for industrial anomaly detection,

    E. Iqbal, S. U. Khan, S. Javed, et al., "Multi-scale feature reconstruction network for industrial anomaly detection," Knowledge-Based Systems, vol. 305, 112650, 2024

    2024

  29. [37]

    Temporal fusion transformers for interpretable multi-horizon time series forecasting,

    B. Lim, S. Ö. Arık, N. Loeff, et al., "Temporal fusion transformers for interpretable multi-horizon time series forecasting," International Journal of Forecasting, vol. 37, no. 4, pp. 1748-1764, 2021

    2021

  30. [38]

    Joint Graph Convolution and Sequential Modeling for Scalable Network Traffic Estimation,

    N. Jiang, W. Zhu, X. Han, W. Huang and Y. Sun, "Joint Graph Convolution and Sequential Modeling for Scalable Network Traffic Estimation," arXiv preprint arXiv:2505.07674, 2025

    2025 arXiv

  31. [39]

    Predictive Modeling of Volatility Using Generative Time-Aware Diffusion Frameworks,

    X. Su, "Predictive Modeling of Volatility Using Generative Time-Aware Diffusion Frameworks," Journal of Computer Technology and Software, vol. 4, no. 5, 2025

    2025

  32. [40]

    Single-Device Human Activity Recognition Based on Spatiotemporal Feature Learning Networks,

    J. Zhan, "Single-Device Human Activity Recognition Based on Spatiotemporal Feature Learning Networks," Transactions on Computational and Scientific Methods, vol. 5, no. 3, 2025

    2025

  33. [41]

    Dynamic operating system scheduling using double DQN: A reinforcement learning approach to task optimization,

    X. Sun, Y. Duan, Y. Deng, F. Guo, G. Cai and Y. Peng, "Dynamic operating system scheduling using double DQN: A reinforcement learning approach to task optimization," Proceedings of the 2025 8th International Conference on Advanced Algorithms and Control Engineering (ICAACE), p...

    2025

  34. [42]

    State-Aware IoT Scheduling Using Deep Q-Networks and Edge-Based Coordination,

    Q. He, C. Liu, J. Zhan, W. Huang and R. Hao, "State-Aware IoT Scheduling Using Deep Q-Networks and Edge-Based Coordination," arXiv preprint arXiv:2504.15577, 2025

    2025 arXiv

  35. [43]

    Autonomous Resource Management in Microservice Systems via Reinforcement Learning,

    Y. Zou, N. Qi, Y. Deng, Z. Xue, M. Gong and W. Zhang, "Autonomous Resource Management in Microservice Systems via Reinforcement Learning," arXiv preprint arXiv:2507.12879, 2025

    2025 arXiv

  36. [44]

    Federated learning for cross-domain data privacy: A distributed approach to secure collaboration,

    Y. Zhang, J. Liu, J. Wang, L. Dai, F. Guo and G. Cai, "Federated learning for cross-domain data privacy: A distributed approach to secure collaboration," arXiv preprint arXiv:2504.00282, 2025

    2025 arXiv

  37. [45]

    Privacy-Enhanced Federated Learning for Distributed Heterogeneous Data,

    T. Kang, H. Yang, L. Dai, X. Hu and J. Du, "Privacy-Enhanced Federated Learning for Distributed Heterogeneous Data," 2025

    2025

  38. [46]

    An Automated Data Mining Framework Using Autoencoders for Feature Extraction and Dimensionality Reduction,

    Y. Liang, X. Li, X. Huang, Z. Zhang and Y. Yao, "An Automated Data Mining Framework Using Autoencoders for Feature Extraction and Dimensionality Reduction," Proceedings of the 2024 4th International Conference on Electronic Information Engineering and Computer Communication (E...

    2024

  39. [47]

    Context-Aware Rule Mining Using a Dynamic Transformer-Based Framework,

    J. Liu, et al., "Context-Aware Rule Mining Using a Dynamic Transformer-Based Framework," Proceedings of the 2025 8th International Conference on Advanced Algorithms and Control Engineering (ICAACE), 2025

    2025

  40. [48]

    Transferable Load Forecasting and Scheduling via Meta- Learned Task Representations,

    T. Yang, "Transferable Load Forecasting and Scheduling via Meta- Learned Task Representations," Journal of Computer Technology and Software, vol. 3, no. 8, 2024

    2024

  41. [49]

    Deep residual learning for image recognition,

    K. He, X. Zhang, S. Ren and J. Sun, "Deep residual learning for image recognition," Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 770-778, 2016

    2016

  42. [50]

    Very deep convolutional networks for large-scale image recognition,

    K. Simonyan and A. Zisserman, "Very deep convolutional networks for large-scale image recognition," arXiv preprint arXiv:1409.1556, 2014

    2014 arXiv

  43. [51]

    A review on the long short-term memory model,

    G. Van Houdt, C. Mosquera and G. Nápoles, "A review on the long short-term memory model," Artificial Intelligence Review, vol. 53, no. 8, pp. 5929-5955, 2020

    2020

  44. [52]

    Learning phrase representations using RNN encoder-decoder for statistical machine translation,

    K. Cho, B. Van Merriënboer, C. Gulcehre, et al., "Learning phrase representations using RNN encoder-decoder for statistical machine translation," arXiv preprint arXiv:1406.1078, 2014

    2014 arXiv

  45. [53]

    Attention is all you need,

    A. Vaswani, N. Shazeer, N. Parmar, et al., "Attention is all you need," Advances in Neural Information Processing Systems, vol. 30, 2017

    2017

  46. [54]

    A multi-information fusion anomaly detection model based on convolutional neural networks and AutoEncoder,

    Z. Zhao, H. Guo and Y. Wang, "A multi-information fusion anomaly detection model based on convolutional neural networks and AutoEncoder," Scientific Reports, vol. 14, no. 1, 16147, 2024

    2024

  47. [55]

    Exathlon: A benchmark for explainable anomaly detection over time series

    V. Jacob, F. Song, A. Stiegler, B. Rad, Y. Diao and N. Tatbul, "Exathlon: A benchmark for explainable anomaly detection over time series", Proceedings of the 2021 PVLDB, vol. 14, no. 12, pp. 2827-2830, 2021

    2021

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.