Pith. sign in

REVIEW 4 major objections 6 minor 1 cited by

Fremer: Lightweight and Effective Frequency Transformer for Workload Forecasting in Cloud Services

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

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

Pith's one-line read The paper argues that a frequency-domain Transformer that aligns input and target spectra with learnable padding, attends over frequency combinations, and filters noise outperforms all tested forecasting models on cloud workload…

desk verdict Worth reading: the learnable padding trick is a genuine new idea and the datasets are useful, but the headline SOTA gains don't survive the controlled comparison and the abstract overclaims. read the letter →

arxiv 2507.12908 v1 pith:MPZU67TJ submitted 2025-07-17 cs.LG

Jiadong Chen , Hengyu Ye , Fuxin Jiang , Xiao He , Tieying Zhang , Jianjun Chen , Xiaofeng Gao This is my paper
classification cs.LG MSC 68T0762M10
keywords workloadforecastingfrequencydomaintransformercloudcomputingtimeseriesresolutionalignmentcomplex-valuedattentionauto-scaling
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

Fremer is a new forecasting model designed for cloud workloads with strong multi-period patterns. The paper's central claim is that working in the frequency domain—predicting the spectrum of the complete input-plus-horizon series—makes forecasting both more accurate and computationally cheaper than time-domain Transformers. The authors report average gains of 5.5% in MSE, 4.7% in MAE, and 8.6% in SMAPE over the best prior models across their datasets and public benchmarks, with large reductions in parameters and compute. They also release four workload datasets from their industrial cloud services and show that in a Kubernetes proactive auto-scaling test the model cuts average latency by 18.78%.

What carries the argument

The core mechanism is Learnable Linear Padding, which pads an input series of length L to length L+T with a learned affine map so that the frequency bins of the input spectrum align with those of the complete input-plus-horizon spectrum. Around this, Fremer builds three other components: Complex-valued Spectrum Attention, which treats single frequency points as semantically weak and computes multi-head attention over linear combinations of frequencies, reducing complexity to O((L')^2/H); Frequency Filters, a low-pass filter that retains trend information while preventing overfitting by keeping it outside the backbone's input, and a high-pass filter that removes the highest-frequency noise; and Frequency Reversible Instance Norm, an adaptation of reversible instance normalization to the frequency domain that makes spectra with different global features comparable.

What would settle it

Re-run the full benchmark with every baseline hyper-parameter-tuned on the same validation split, using the same normalization, hardware, and training configuration, and re-run the general-forecasting comparisons in the same environment; if Fremer's average MSE advantage over the best baseline falls to near zero or reverses, the paper's central claim of consistent superiority fails.

Watch

Extended reading notes

Core claim

Fremer's central claim is that forecasting in the frequency domain, applied to the spectrum of the input padded to the full forecast horizon, is both more accurate and more efficient than existing time-domain and frequency-domain approaches for cloud workload series. The paper identifies a previously unaddressed problem called frequency resolution misalignment: because the discrete Fourier transform samples frequencies at intervals of 1/L for input length L, the true periodic frequencies of the complete series are often not present in the input spectrum, so matching input and target spectra is inherently difficult. Fremer solves this with a learnable linear layer that pads the input to length L+T, aligns the frequency grids, then applies complex-valued attention over combinations of frequencies, plus low-pass and high-pass filters that retain trend and discard noise. On seven workload datasets and several public benchmarks, the authors report consistent improvements over state-of-the-art models, with average reductions of 5.5% in MSE, 4.7% in MAE, and 8.6% in SMAPE, while cutting parameter counts by roughly an order of magnitude and improving inference throughput.

Load-bearing premise

The load-bearing premise is that the comparison against baselines is fair: Fremer gets a limited hyper-parameter search focused on the frequency filter threshold, whereas baselines run with default settings from the benchmark platform, and the general-forecasting results are taken directly from the platform's paper rather than re-run under identical conditions. If that premise fails, the claim of consistent superiority over all state-of-the-art models is weakened, though Fremer may still be competitive and efficient.

Editorial extensions

If this is right

  • If the central claim holds, Fremer offers a better accuracy-efficiency tradeoff than existing Transformer forecasters on cloud workload series, making it feasible for large-scale deployments where hundreds of thousands of forecasts run per hour.
  • Channel-independent frequency attention transfers to unseen instances and new datasets, so the model can be used as a zero-shot forecasting backbone without retraining per instance.
  • In a Kubernetes proactive auto-scaling setup, using Fremer's forecasts reduces average latency by 18.78% and average pod consumption by 2.35% compared with a PatchTST-based autoscaler.
  • On general benchmarks with strong periodicity (Traffic, Electricity, PEMS04), Fremer outperforms baselines, suggesting the frequency-domain design generalizes beyond cloud workloads.
  • The frequency filters provide a simple, tunable mechanism to control overfitting: the low-pass threshold directly trades training fit against validation and test error.

Reading between the lines

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

  • Editorial extension: the frequency-resolution misalignment problem likely affects other frequency-domain forecasters that compare input and output spectra at different resolutions; the learnable linear padding idea could be lifted and inserted into those models as a general module.
  • Editorial extension: the finding that attention over frequency combinations works better than attention over individual frequency points suggests that harmonic structure, not individual peaks, is the right inductive bias for periodic workloads; this could be tested on non-cloud series with chirp-like or drifting periodicities.
  • Editorial extension: the Kubernetes test uses a single function instance under replay, so the claimed 18.78% latency gain may not directly scale to mixed multi-tenant clusters; replicating the test across the full FaaS dataset and multiple instance types would tell how robust the benefit is.
  • Editorial extension: because the low-pass filter threshold strongly influences overfitting, the threshold could be adapted per series or per dataset using validation loss, turning Fremer into a self-regularizing forecaster without extra training cost.
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 Fremer, an encoder-only Transformer that forecasts in the frequency domain. It pads the input with a learnable linear layer (LLP) to align the DFT frequency grids of input and complete series, applies low- and high-pass frequency filters, and uses a complex-valued spectrum attention (CSA) mechanism over frequency combinations. The model is evaluated on four new ByteDance workload datasets, three Materna workload traces, and eleven TFB general-forecasting datasets, with additional efficiency, ablation, transfer, and Kubernetes HPA experiments. The authors claim average improvements of 5.5% MSE, 4.7% MAE, and 8.6% SMAPE over SOTA models, while using 0.57M parameters and lower computational cost.

Significance. If the claims are established, Fremer would be a valuable accuracy-efficiency tradeoff for cloud workload forecasting, and the open-sourced ByteDance datasets would be a useful community resource. The paper also gives a concrete treatment of DFT frequency-resolution misalignment, which is a real and underappreciated issue in frequency-domain forecasting. However, the headline superiority claim is currently not supported by a single controlled protocol: the workload-forecasting results in Table 3 show smaller gains and two datasets where Fremer is not best, while the larger gains in the abstract appear to rely on Table 7 whose baselines are copied from the TFB paper. The core CSA definition also has notational problems that obscure whether the mechanism actually performs attention over frequency combinations.

major comments (4)
  1. [Abstract and §5.2] The abstract claims average improvements of 5.5% in MSE, 4.7% in MAE, and 8.6% in SMAPE over SOTA models, but the controlled workload comparison in §5.2 reports only 2.9%, 2.5%, and 3.5%. Recomputing from Table 3 gives roughly 2.8% MSE, 2.5% MAE, and 3.8% SMAPE against the best baseline per dataset, and Fremer is not the best on MT1 MSE (15.408 vs. iTransformer 15.205) or MT2 MSE (6.543 vs. iTransformer/Crossformer 6.467). The larger abstract numbers must come from Table 7, whose baselines are 'directly referenced from the TFB paper' rather than measured under Fremer's protocol. The authors should either report the headline numbers from the controlled workload experiment or clearly label the abstract claim as referring to the uncontrolled general-forecasting setting.
  2. [§5.6 and Table 7] Table 7 cannot support the claim that Fremer 'outperforms all baseline models' on general forecasting. Baseline results are copied from TFB, so they were produced with different normalization, input lengths, hardware, and hyperparameter settings, while §5.1 states that Fremer underwent its own hyperparameter search. This is not a controlled comparison. Moreover, Table 7 itself shows counterexamples to the text's claim of superiority on strongly periodic datasets: on PEMS08, Fremer's MSE is 0.288 versus 0.213 for PatchTST and 0.176 for Crossformer. The authors should rerun the baselines under the identical protocol used for Fremer, or restrict the general-forecasting claims to descriptive observations.
  3. [§4.3, Eqs. (4)-(5)] The definition of Complex-valued Spectrum Attention is dimensionally inconsistent as written. F_c is defined as an element of C^{L'}, and Q_h = F_c^T W_Q^h with W_Q^h ∈ C^{L'×l} gives Q_h ∈ C^{1×l}; the same holds for K_h and V_h. Then Q_h K_h^T is a 1×1 matrix, so the softmax is trivial and the 'attention' output is just V_h scaled by 1, with no mixing across frequency combinations. The later statement in §4.4 that Q,K,V ∈ C^{1×L'} has the same problem. The authors should provide the correct tensor dimensions (e.g., whether F_c is a matrix of combinations, or whether Q,K,V are L'×d matrices) and show how the claimed O((L')^2/H) complexity arises. As written, the central architectural mechanism is not well defined.
  4. [§5.1 and §5.6] There is an internal contradiction about the experimental protocol. §5.1 says 'We implement Fremer and all baseline models utilizing the PyTorch framework' and that 'a uniform configuration is applied across all models,' while §5.6 says all baseline results for general forecasting are 'directly referenced from the TFB paper.' These two statements cannot both describe the same evaluation. The authors should state explicitly which baseline numbers were measured and which were imported, and ensure that the headline claims are based only on the measured comparisons.
minor comments (6)
  1. [Global] There are several typos and formatting errors, including 'Univerisity' in the author affiliation, 'Alighment' in the §4.2 heading, and the malformed table entry '11.6431' in Table 3. A careful proofreading pass is needed.
  2. [§4.4] The sentence 'The dot-product Attention mechanism has a complexity of O(L)' contradicts the standard O(L^2) complexity and the paper's own later formula; this should be corrected to O(L^2).
  3. [§5.3.1 / Figure 8] The ablation study is reported only as bar charts without numerical values or error bars, making it hard for readers to assess the magnitude of each component's contribution. Reporting the exact numbers or a small table would improve reproducibility.
  4. [§5.7 / Table 8] The Kubernetes HPA experiment uses a single function instance from the FaaS dataset, so the 18.78% latency improvement and 2.35% resource reduction have no statistical characterization. The authors should state whether these results are representative and, ideally, report variability across multiple instances.
  5. [§4.1] The claim of being 'the first to address' frequency misalignment should be softened or supported by a discussion of related spectral interpolation and zero-padding techniques in the signal-processing and forecasting literature.
  6. [Table 6] The efficiency comparison includes only Transformer-based baselines, not all SOTA models such as FITS or DLinear. Since the abstract claims improved efficiency relative to SOTA overall, the table caption or text should clarify that the comparison is restricted to Transformer baselines.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: Fremer is an empirical neural forecasting model whose components are learned on held-out data; no claimed derivation reduces to its inputs or to self-citations.

full rationale

Fremer's derivation chain is empirical: an input series is padded with a learnable linear map, transformed by rFFT, filtered, normalized, processed by complex-valued spectrum attention, and transformed back with irFFT, with the last T points compared to held-out ground truth. None of the quoted equations defines the target as an input (Eqs. 1-5 are standard DFT/attention operations), and no learned component is a renamed measurement of the forecast target. The LLP, F-filter, F-RIN, and CSA parameters are trained with MSE loss on a 7:1:2 split; the limited hyper-parameter search over the filter threshold is ordinary model selection, not a fitted parameter presented as a prediction. The paper's only self-citation (ref. [12], OneShotSTL, by co-author Xiao He) supports a related-work remark about STL and is not load-bearing. The 'directly referenced from the TFB paper' baseline numbers in Section 5.6 raise an evaluation-fairness question for the headline SOTA percentages, and the abstract's 5.5/4.7/8.6 improvement figures are not reproduced by the controlled Section 5.2 comparison (2.9/2.5/3.5); these are consistency and attribution issues, not circular reductions. No circular step can be exhibited, so the score is 0.

Assumptions & free parameters 4 free parameters · 7 assumptions · 0 invented entities

Fremer introduces several tunable hyperparameters (filter thresholds, combination count, head count) that are selected via sensitivity analysis and directly affect performance; the exact tuned values are not reported. The architecture assumes frequency-domain concentration and channel independence, both empirical statements validated only on the tested datasets. No new physical entities are introduced.

free parameters (4)
  • LPF threshold = not reported (default highest 1% of frequencies, tuned per dataset)
    The low-pass filter threshold sets which high-frequency components are zeroed. The paper says the default is the highest 1% but that a limited hyperparameter search focused on this threshold; exact chosen values are not given.
  • HPF threshold = not reported (default lowest 3% of frequencies, tuned per dataset)
    The high-pass filter threshold selects which low-frequency trend components are excluded from the backbone. The default is the lowest 3%, but tuned values are not reported.
  • Frequency combination count L' = L/5 (e.g., 144 for L=720)
    The number of frequency combinations after linear projection. Sensitivity analysis in Figure 10(a) shows L/5 is optimal for L=720; it is a hyperparameter that directly controls complexity.
  • Attention heads H = 8
    The number of heads in complex-valued attention. Figure 10(b) shows H=8 balances performance and efficiency; larger H gives diminishing returns.
assumptions (7)
  • standard math DFT/FFT pair and iDFT reconstruction
    The method relies on rFFT and irFFT to convert between time and frequency domains, and on the Hermitian symmetry of real signals.
  • domain assumption Workload series exhibit strong periodic patterns (hourly, daily, weekly)
    The motivation in the introduction and Figure 1 asserts that cloud workloads have complicated periodic structure, which justifies the frequency-domain approach.
  • domain assumption Frequency-domain information is concentrated in a few key frequencies and appears in combinations (harmonics)
    Section 4.3 argues a single frequency point lacks semantic meaning, so attention over frequency combinations is needed; this is an empirical assumption about time series spectra.
  • domain assumption Channel independence is effective for workload data
    The model processes each variate separately, citing PatchTST's channel independence; Table 3 results support this but it remains an assumption for new datasets.
  • domain assumption RevIN (instance normalization) transfers to the frequency domain
    Section 4.3 claims F-RIN helps normalize spectra with different global features, a claim tested only in the ablation (w/o F-RIN) on four datasets.
  • ad hoc to paper The padded input spectrum can be transformed to the complete-series spectrum via attention and filters
    The entire Fremer architecture relies on the learnability of this mapping; it is a design premise rather than a proven theorem.
  • domain assumption Workload volume scales linearly with resource consumption
    The Kubernetes HPA test in Section 5.7 bases scaling on the assumption that workload volume has a linear relationship with resource consumption.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Fremer: Lightweight and Effective Frequency Transformer for Workload Forecasting in Cloud Services." pith.science (2026). https://pith.science/paper/MPZU67TJ

@misc{pith2026250712908,
  author       = {Pith},
  title        = {Pith review of: Fremer: Lightweight and Effective Frequency Transformer for Workload Forecasting in Cloud Services},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/MPZU67TJ}},
  note         = {Machine review of arXiv:2507.12908}
}
read the original abstract

Workload forecasting is pivotal in cloud service applications, such as auto-scaling and scheduling, with profound implications for operational efficiency. Although Transformer-based forecasting models have demonstrated remarkable success in general tasks, their computational efficiency often falls short of the stringent requirements in large-scale cloud environments. Given that most workload series exhibit complicated periodic patterns, addressing these challenges in the frequency domain offers substantial advantages. To this end, we propose Fremer, an efficient and effective deep forecasting model. Fremer fulfills three critical requirements: it demonstrates superior efficiency, outperforming most Transformer-based forecasting models; it achieves exceptional accuracy, surpassing all state-of-the-art (SOTA) models in workload forecasting; and it exhibits robust performance for multi-period series. Furthermore, we collect and open-source four high-quality, open-source workload datasets derived from ByteDance's cloud services, encompassing workload data from thousands of computing instances. Extensive experiments on both our proprietary datasets and public benchmarks demonstrate that Fremer consistently outperforms baseline models, achieving average improvements of 5.5% in MSE, 4.7% in MAE, and 8.6% in SMAPE over SOTA models, while simultaneously reducing parameter scale and computational costs. Additionally, in a proactive auto-scaling test based on Kubernetes, Fremer improves average latency by 18.78% and reduces resource consumption by 2.35%, underscoring its practical efficacy in real-world applications.

Figures

Figures reproduced from arXiv: 2507.12908 by the authors.

Figure 1
Figure 1. Workload from cloud services in ByteDance exhibit [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Comparison of throughput and prediction error [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 3
Figure 3. Model Architecture of Fremer. As illustrated in [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (8 more)
Figure 4
Figure 4. Figure 4: Frequency Resolution Mis-alignment. The original workload series is sourced from PaaS dataset. It could be observed [PITH_FULL_IMAGE:figures/full_fig_p005_4.png]
Figure 5
Figure 5. Figure 5: Frequency Filter is used to retain low-frequency [PITH_FULL_IMAGE:figures/full_fig_p005_5.png]
Figure 6
Figure 6. Figure 6: Comparison between Frequency Enhanced Attention (FEA) proposed by FEDformer [61]) and our proposed CSA. [PITH_FULL_IMAGE:figures/full_fig_p006_6.png]
Figure 7
Figure 7. Figure 7: Visualization of Workload Forecasting Results on Distinct Datasets. [PITH_FULL_IMAGE:figures/full_fig_p009_7.png]
Figure 8
Figure 8. Figure 8: Ablation Study Results. We highlight the contributions of four key modules of [PITH_FULL_IMAGE:figures/full_fig_p009_8.png]
Figure 9
Figure 9. Figure 9: Frequency Filter Analysis Results. Figure (a) and [PITH_FULL_IMAGE:figures/full_fig_p009_9.png]
Figure 12
Figure 12. Figure 12: Multi-head CSA. The figure shows the mapping [PITH_FULL_IMAGE:figures/full_fig_p010_12.png]
Figure 11
Figure 11. Figure 11: The sensitivity to 𝐿 and 𝑇 of Fremer. The three axes respectively represent the input length 𝐿, the forecast horizon 𝑇 , and the SMAPE. Amplitude Frequency Head 1 Head 2 Head 3 Head 4 Head 5 Head 6 Head 7 Head 8 ... Frequency Points Frequency Combinations Attention Sc…

Discussion (0). Sign in 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. SWIFT: Spatio-temporal Wavelet Integrated Forecasting Framework for Workload Traces

    cs.DC 2026-05 conditional novelty 6.0 of 10

    SWIFT forecasts cloud workloads at SOTA accuracy with linear complexity by combining learnable cascaded wavelet convolutions and sequential inter/intra-variable mixing.

Reference graph

Works this paper leans on

66 extracted references · 41 canonical work pages · cited by 1 Pith paper

  1. [1]

    Yanal Alahmad, Tariq Daradkeh, and Anjali Agarwal. 2021. Proactive failure- aware task scheduling framework for cloud computing.IEEE Access9 (2021), 106152–106168

    work page 2021

  2. [2]

    Alexandru-Florian Antonescu and Torsten Braun. 2016. Simulation of SLA-based VM-scaling algorithms for cloud-distributed applications.Future Generation computer systems54 (2016), 260–273

    work page 2016

  3. [3]

    Shaojie Bai, J Zico Kolter, and Vladlen Koltun. 2018. An empirical evaluation of generic convolutional and recurrent networks for sequence modeling.arXiv preprint arXiv:1803.01271(2018)

    arXiv 2018

  4. [4]

    Masoud Barati and Saeed Sharifian. 2015. A hybrid heuristic-based tuned support vector regression model for cloud load prediction.The Journal of Supercomputing 71 (2015), 4235–4259

    work page 2015

  5. [5]

    George EP Box and Gwilym M Jenkins. 1968. Some recent advances in forecasting and control.Journal of the Royal Statistical Society. Series C (Applied Statistics) 17, 2 (1968), 91–109

    work page 1968

  6. [6]

    Rodrigo N Calheiros, Enayat Masoumi, Rajiv Ranjan, and Rajkumar Buyya. 2014. Workload prediction using ARIMA model and its impact on cloud applications’ QoS.IEEE transactions on cloud computing3, 4 (2014), 449–458

    work page 2014

  7. [7]

    Olivares, Boris N

    Cristian Challu, Kin G. Olivares, Boris N. Oreshkin, Federico Garza, Max Mergenthaler-Canseco, and Artur Dubrawski. 2022. N-HiTS: Neural Hierar- chical Interpolation for Time Series Forecasting. arXiv:2201.12886 [cs.LG]

    arXiv 2022

  8. [8]

    Junyoung Chung, Caglar Gulcehre, KyungHyun Cho, and Yoshua Bengio. 2014. Empirical evaluation of gated recurrent neural networks on sequence modeling. arXiv preprint arXiv:1412.3555(2014)

    arXiv 2014

Show all 66 references
  1. [9]

    Tao Dai, Beiliang Wu, Peiyuan Liu, Naiqi Li, Jigang Bao, Yong Jiang, and Shu-Tao Xia. 2024. Periodicity decoupling framework for long-term series forecasting. In International Conference on Learning Representations

    2024

  2. [10]

    Zhihan Gao, Xingjian Shi, Hao Wang, Yi Zhu, Yuyang Bernie Wang, Mu Li, and Dit-Yan Yeung. 2022. Earthformer: Exploring space-time transformers for earth system forecasting.Advances in Neural Information Processing Systems35 (2022), 25390–25403

    2022

  3. [11]

    Ehsan Golshani and Mehrdad Ashtiani. 2021. Proactive auto-scaling for cloud environments using temporal convolutional neural networks.J. Parallel and Distrib. Comput.154 (2021), 119–141

    2021

  4. [12]

    Xiao He, Ye Li, Jian Tan, Bin Wu, and Feifei Li. 2023. OneShotSTL: One-Shot Seasonal-Trend Decomposition For Online Time Series Anomaly Detection And Forecasting.Proc. VLDB Endow.16, 6 (2023), 1399–1412. https://doi.org/10.14778/ 3583140.3583155

    2023

  5. [13]

    Charles C Holt. 2004. Forecasting seasonals and trends by exponentially weighted moving averages.International journal of forecasting20, 1 (2004), 5–10

    2004

  6. [14]

    Maowei Jiang, Pengyu Zeng, Kai Wang, Huan Liu, Wenbo Chen, and Haoran Liu. 2023. FECAM: Frequency enhanced channel attention mechanism for time series forecasting.Advanced Engineering Informatics58 (2023), 102158

    2023

  7. [15]

    Md Ebtidaul Karim, Mirza Mohd Shahriar Maswood, Sunanda Das, and Abdul- lah G Alharbi. 2021. BHyPreC: a novel Bi-LSTM based hybrid recurrent neural network model to predict the CPU workload of cloud virtual machine.IEEE Access9 (2021), 131476–131495

    2021

  8. [16]

    Tahseen Khan, Wenhong Tian, Shashikant Ilager, and Rajkumar Buyya. 2022. Workload forecasting and energy state estimation in cloud data centres: ML- centric approach.Future Generation Computer Systems128 (2022), 320–332

    2022

  9. [17]

    Reihaneh Khorsand, Mostafa Ghobaei-Arani, and Mohammadreza Ramezanpour

  10. [18]

    Taesung Kim, Jinhee Kim, Yunwon Tae, Cheonbok Park, Jang-Ho Choi, and Jaegul Choo. 2021. Reversible instance normalization for accurate time-series forecasting against distribution shift. InInternational Conference on Learning Representations

    2021

  11. [19]

    Diederik P Kingma and Jimmy Ba. 2014. Adam: A method for stochastic opti- mization.arXiv preprint arXiv:1412.6980(2014)

    2014 arXiv

  12. [20]

    Nikita Kitaev, Łukasz Kaiser, and Anselm Levskaya. 2020. Reformer: The efficient transformer.arXiv preprint arXiv:2001.04451(2020)

    2020 arXiv

  13. [21]

    Anoop S Kumar and Somnath Mazumdar. 2016. Forecasting HPC workload using ARMA models and SSA. In2016 International conference on information technology (ICIT). IEEE, 294–297

    2016

  14. [22]

    Guokun Lai, Wei-Cheng Chang, Yiming Yang, and Hanxiao Liu. 2018. Modeling long-and short-term temporal patterns with deep neural networks. InThe 41st international ACM SIGIR conference on research & development in information retrieval. 95–104

    2018

  15. [23]

    Patricio Lamas and Erik Demeulemeester. 2016. A purely proactive schedul- ing procedure for the resource-constrained project scheduling problem with stochastic activity durations.Journal of Scheduling19 (2016), 409–428

    2016

  16. [24]

    Habte Lejebo Leka, Zhang Fengli, Ayantu Tesfaye Kenea, Abebe Tamrat Tegene, Peter Atandoh, and Negalign Wake Hundera. 2021. A hybrid cnn-lstm model for virtual machine workload forecasting in cloud data center. InIEEE Interna- tional Computer Conference on Wavelet Active Media...

    2021

  17. [25]

    Shengming Li, Ying Wang, Xuesong Qiu, Deyuan Wang, and Lijun Wang. 2013. A workload prediction-based multi-vm provisioning mechanism in cloud comput- ing. InAsia-Pacific Network Operations and Management Symposium (APNOMS). IEEE, 1–6

    2013

  18. [26]

    Yuxuan Liang, Yutong Xia, Songyu Ke, Yiwei Wang, Qingsong Wen, Junbo Zhang, Yu Zheng, and Roger Zimmermann. 2023. Airformer: Predicting nationwide air quality in china with transformers. InAAAI conference on artificial intelligence, Vol. 37. 14329–14337

    2023

  19. [27]

    Shengsheng Lin, Weiwei Lin, Wentai Wu, Haojun Chen, and Junjie Yang. 2024. Sparsetsf: Modeling long-term time series forecasting with 1k parameters.arXiv preprint arXiv:2405.00946(2024)

    2024 arXiv

  20. [28]

    Minhao Liu, Ailing Zeng, Muxi Chen, Zhijian Xu, Qiuxia Lai, Lingna Ma, and Qiang Xu. 2022. Scinet: Time series modeling and forecasting with sample convolution and interaction.Advances in Neural Information Processing Systems 35 (2022), 5816–5828

    2022

  21. [29]

    Shizhan Liu, Hang Yu, Cong Liao, Jianguo Li, Weiyao Lin, Alex X Liu, and Schahram Dustdar. 2022. Pyraformer: Low-complexity pyramidal attention for long-range time series modeling and forecasting. InInternational Conference on Learning Representations

    2022

  22. [30]

    Yong Liu, Tengge Hu, Haoran Zhang, Haixu Wu, Shiyu Wang, Lintao Ma, and Mingsheng Long. 2023. itransformer: Inverted transformers are effective for time series forecasting.arXiv preprint arXiv:2310.06625(2023)

    2023 arXiv

  23. [31]

    Federico Lombardi, Andrea Muti, Leonardo Aniello, Roberto Baldoni, Silvia Bonomi, and Leonardo Querzoni. 2019. Pascal: An architecture for proactive auto-scaling of distributed services.Future Generation Computer Systems98 (2019), 342–361

    2019

  24. [32]

    Laura R Moore, Kathryn Bean, and Tariq Ellahi. 2013. Transforming reactive auto-scaling into proactive auto-scaling. InProceedings of the 3rd International Workshop on Cloud Data and Platforms. 7–12

    2013

  25. [33]

    Hoang Minh Nguyen, Sungpil Woo, Janggwan Im, Taejoon Jun, and Daeyoung Kim. 2016. A workload prediction approach using models stacking based on recur- rent neural network and autoencoder. In2016 IEEE 18th International Conference on High Performance Computing and Communication...

    2016

  26. [34]

    Zelin Ni, Hang Yu, Shizhan Liu, Jianguo Li, and Weiyao Lin. 2023. Basisformer: Attention-based time series forecasting with learnable and interpretable basis. Advances in Neural Information Processing Systems36 (2023), 71222–71241

    2023

  27. [35]

    Yuqi Nie, Nam H Nguyen, Phanwadee Sinthong, and Jayant Kalagnanam. 2022. A Time Series is Worth 64 Words: Long-term Forecasting with Transformers. In The Eleventh International Conference on Learning Representations

    2022

  28. [36]

    Boris N Oreshkin, Dmitri Carpov, Nicolas Chapados, and Yoshua Bengio. 2019. N- BEATS: Neural basis expansion analysis for interpretable time series forecasting. arXiv preprint arXiv:1905.10437(2019)

    2019 arXiv

  29. [37]

    Adam Paszke, Sam Gross, Francisco Massa, Adam Lerer, James Bradbury, Gregory Chanan, Trevor Killeen, Zeming Lin, Natalia Gimelshein, Luca Antiga, et al. 2019. Pytorch: An imperative style, high-performance deep learning library.Advances in neural information processing systems...

    2019

  30. [38]

    Xihao Piao, Zheng Chen, Taichi Murayama, Yasuko Matsubara, and Yasushi Sakurai. 2024. Fredformer: Frequency debiased transformer for time series forecasting. InACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD). 2400–2410

    2024

  31. [39]

    Jensen, Zhenli Sheng, and Bin Yang

    Xiangfei Qiu, Jilin Hu, Lekui Zhou, Xingjian Wu, Junyang Du, Buang Zhang, Chenjuan Guo, Aoying Zhou, Christian S. Jensen, Zhenli Sheng, and Bin Yang

  32. [40]

    Bane Raman Raghunath and B Annappa. 2015. Virtual machine migration triggering using application workload prediction.Procedia Computer Science54 (2015), 167–176

    2015

  33. [41]

    Dymitr Ruta, Ling Cen, and Quang Hieu Vu. 2020. Deep bi-directional LSTM networks for device workload forecasting. In2020 15th Conference on Computer Science and Information Systems (FedCSIS). IEEE, 115–118

    2020

  34. [42]

    Deepika Saxena and Ashutosh Kumar Singh. 2021. Workload forecasting and resource management models based on machine learning for cloud computing environments.arXiv preprint arXiv:2106.15112(2021)

    2021 arXiv

  35. [43]

    RS Shariffdeen, DTSP Munasinghe, HS Bhathiya, UKJU Bandara, and HMN Dilum Bandara. 2016. Adaptive workload prediction for proactive auto scaling in PaaS systems. InIEEE International Conference on Cloud Computing Technologies and Applications (CloudTech). 22–29

    2016

  36. [44]

    Binbin Song, Yao Yu, Yu Zhou, Ziqiang Wang, and Sidan Du. 2018. Host load prediction with long short-term memory in cloud computing.The Journal of Supercomputing74 (2018), 6554–6568

    2018

  37. [45]

    Fan-Keng Sun and Duane S Boning. 2022. FreDo: Frequency Domain-based Long-Term Time Series Forecasting.arXiv e-prints(2022), arXiv–2205

    2022

  38. [46]

    Sundararajan

    D. Sundararajan. 2025.The Discrete Fourier Transform. Springer Nature Singapore, Singapore, 37–67. https://doi.org/10.1007/978-981-96-1078-5_2

    2025 doi

  39. [47]

    2018.Energy efficient dynamic virtual machine allocation with CPU usage prediction in cloud datacenters

    Gökalp Urul. 2018.Energy efficient dynamic virtual machine allocation with CPU usage prediction in cloud datacenters. Master’s thesis. Bilkent Universitesi Jiadong Chen1,3,∗, Hengyu Ye1,∗, Fuxin Jiang2, Xiao He2, Tieying Zhang2, Jianjun Chen2, Xiaofeng Gao1,† (Turkey)

    2018

  40. [48]

    Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N Gomez, Łukasz Kaiser, and Illia Polosukhin. 2017. Attention is all you need.Advances in neural information processing systems30 (2017)

    2017

  41. [49]

    Huiqiang Wang, Jian Peng, Feihu Huang, Jince Wang, Junhui Chen, and Yifei Xiao

  42. [50]

    Gerald Woo, Chenghao Liu, Doyen Sahoo, Akshat Kumar, and Steven Hoi. 2022. Etsformer: Exponential smoothing transformers for time-series forecasting.arXiv preprint arXiv:2202.01381(2022)

    2022 arXiv

  43. [51]

    Haixu Wu, Tengge Hu, Yong Liu, Hang Zhou, Jianmin Wang, and Mingsheng Long. 2022. TimesNet: Temporal 2D-Variation Modeling for General Time Series Analysis. InThe Eleventh International Conference on Learning Representations

    2022

  44. [52]

    Haixu Wu, Jiehui Xu, Jianmin Wang, and Mingsheng Long. 2021. Autoformer: De- composition transformers with auto-correlation for long-term series forecasting. Advances in Neural Information Processing Systems34 (2021), 22419–22430

    2021

  45. [53]

    Zhijian Xu, Ailing Zeng, and Qiang Xu. 2024. FITS: Modeling Time Series with 10𝑘Parameters. arXiv:2307.03756 [cs.LG]

    2024 arXiv

  46. [54]

    Jingqi Yang, Chuanchang Liu, Yanlei Shang, Zexiang Mao, and Junliang Chen

  47. [55]

    Kun Yi, Jingru Fei, Qi Zhang, Hui He, Shufeng Hao, Defu Lian, and Wei Fan. 2024. Filternet: Harnessing frequency filters for time series forecasting.Advances in Neural Information Processing Systems (NeurIPS)37 (2024), 55115–55140

    2024

  48. [56]

    Kun Yi, Qi Zhang, Wei Fan, Shoujin Wang, Pengyang Wang, Hui He, Ning An, Defu Lian, Longbing Cao, and Zhendong Niu. 2023. Frequency-domain MLPs are More Effective Learners in Time Series Forecasting. InThirty-seventh Conference on Neural Information Processing Systems

    2023

  49. [57]

    Ailing Zeng, Muxi Chen, Lei Zhang, and Qiang Xu. 2023. Are transformers effective for time series forecasting?. InProceedings of the AAAI conference on artificial intelligence, Vol. 37. 11121–11128

    2023

  50. [58]

    Yunhao Zhang and Junchi Yan. 2022. Crossformer: Transformer utilizing cross- dimension dependency for multivariate time series forecasting. InThe Eleventh International Conference on Learning Representations

    2022

  51. [59]

    Haoyi Zhou, Shanghang Zhang, Jieqi Peng, Shuai Zhang, Jianxin Li, Hui Xiong, and Wancai Zhang. 2021. Informer: Beyond efficient transformer for long se- quence time-series forecasting. InProceedings of the AAAI conference on artificial intelligence, Vol. 35. 11106–11115

    2021

  52. [60]

    Tian Zhou, Ziqing Ma, Qingsong Wen, Liang Sun, Tao Yao, Wotao Yin, Rong Jin, et al. 2022. Film: Frequency improved legendre memory model for long-term time series forecasting.Advances in Neural Information Processing Systems35 (2022), 12677–12690

    2022

  53. [61]

    Tian Zhou, Ziqing Ma, Qingsong Wen, Xue Wang, Liang Sun, and Rong Jin

  54. [2013]

    In2013 IEEE Sixth International Conference on Cloud Computing

    Workload predicting-based automatic scaling in service clouds. In2013 IEEE Sixth International Conference on Cloud Computing. IEEE, 810–815

  55. [2018]

    WITHDRAWN: a fuzzy auto-scaling approach using workload prediction for MMOG application in a cloud environment

  56. [2022]

    InInternational Conference on Machine Learning

    Fedformer: Frequency enhanced decomposed transformer for long-term series forecasting. InInternational Conference on Machine Learning. PMLR, 27268– 27286

  57. [2023]

    InThe eleventh international conference on learning representations

    Micn: Multi-scale local and global context modeling for long-term series forecasting. InThe eleventh international conference on learning representations

  58. [2024]

    VLDB Endow.17, 9 (2024), 2363–2377

    TFB: Towards Comprehensive and Fair Benchmarking of Time Series Forecasting Methods.Proc. VLDB Endow.17, 9 (2024), 2363–2377

    2024

Pith tools

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