CASA: CNN Autoencoder-based Score Attention for Efficient Multivariate Long-term Time-series Forecasting

Hyekyung Yoon; Minhyuk Lee; Myungjoo Kang

REVIEW 4 major objections 5 minor 25 references

CASA: CNN Autoencoder-based Score Attention for Efficient Multivariate Long-term Time-series Forecasting

T0 review · 4 major / 5 minor · reviewed 2026-08-16 · deepseek-v4-flash

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

Pith's one-line read This paper claims that replacing Transformer attention scores with a CNN autoencoder makes multivariate time-series forecasting faster, lighter, and more accurate.

desk verdict A useful empirical module whose theoretical motivation is wrong; the paper deserves a serious referee after reframing. read the letter →

arxiv 2505.02011 v1 pith:URJD6M54 submitted 2025-05-04 cs.LG cs.AI

Minhyuk Lee , HyeKyung Yoon , MyungJoo Kang This is my paper · ORCID

classification cs.LGcs.AI

keywords multivariatelong-termtimeseriesforecastingCNNautoencoderattentionscoreapproximationchannel-wisetokenizationcross-dimensionalinteractionslinearcomplexityefficientTransformer

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 aims to establish that the attention score matrix $QK^\top/\sqrt{d_k}$ in Transformers can be replaced by a CNN autoencoder that learns the score map directly, and that this substitution improves both efficiency and accuracy for multivariate long-term time-series forecasting. The authors argue that standard query and key embeddings are variate-independent under channel-wise tokenization, and time-independent under point/patch tokenization, so the attention score never uses correlations between variates; the CNN score network restores that information by treating variates as channels. On eight real-world datasets they report 54 of 64 best MSE/MAE scores, up to 77.7% memory reduction, and 44.0% faster inference, with the module working across point-wise, patch-wise, and channel-wise Transformers. The practical stakes are that attention is the main computational bottleneck in forecasting Transformers, and a drop-in linear-complexity replacement that also improves accuracy would remove a major cost barrier.

What carries the argument

The central object is the score network: a 1D CNN autoencoder with an inverted bottleneck, used in place of $QK^\top/\sqrt{d_k}$. It takes the embedded feature $Z_i \in \mathbb{R}^{N\times D}$ with variates as channels, expands it through convolution into a high-dimensional latent space, then compresses back to an attention score map; softmax of that map is multiplied element-wise by the value embedding. This carries the argument because it makes query/key construction variate-dependent and keeps the overall complexity $O(NL + NLD + ND^2 + NkD^2 + NDH)$, which scales linearly in $N$, $L$, and $H$ when the hidden dimension $D$ and kernel size $k$ are treated as constants.

What would settle it

Run the Traffic benchmark at $L=96$, $H=720$, batch size 16 and measure CASA's memory and latency against iTransformer and SOFTS on identical hardware; if the reduction is only the roughly 2% gap shown in Table 1 rather than the claimed 77.7%, the central efficiency claim fails. Separately, if CASA's advantage vanishes on a dataset with weak inter-variate correlations, the cross-dimensional mechanism is not the cause.

Watch

Extended reading notes

Core claim

CASA treats each variate as a channel and replaces the affine query/key embeddings of self-attention with a 1D CNN autoencoder that approximates $QK^\top/\sqrt{d_k}$; the attention output is then softmax of the learned score, multiplied element-wise by the value embedding. The paper proves that with channel-wise tokenization the standard query and key embeddings are variate-independent (Proposition 1), and with point-wise or patch-wise tokenization they are time-independent (Proposition 2), so conventional attention scores are built without cross-variate information. The CNN score network, built as an inverted-bottleneck autoencoder, expands all variates into a high-dimensional latent space and then compresses channels, making score construction variate-dependent at claimed linear cost in $N$, $L$, and $H$. In the authors' experiments, swapping self-attention for CASA improved 40 of 42 metrics across vanilla Transformer, PatchTST, and iTransformer, and the full CASA model ranked first in 54 of 64 metrics across eight datasets.

Load-bearing premise

The efficiency story assumes the hidden dimension and kernel width are small fixed constants, but in the reported experiments those constants are large enough that quadratic terms can dominate, and the headline 77.7% memory saving is not tied to a clear baseline in the provided tables.

Editorial extensions

If this is right

Transformer encoders for multivariate series can be made linear in variate count, input length, and horizon without sacrificing accuracy; the paper reports first place in 54 of 64 metrics across eight datasets.
The module is a drop-in replacement: swapping self-attention for CASA improved 40 of 42 evaluated metrics across vanilla Transformer, PatchTST, and iTransformer, so no change in tokenization is required.
On high-variate data such as Traffic with 862 channels, CASA's predicted inter-variate correlation matrix matches the ground truth more closely than the compared baselines, indicating the score network preserves cross-variate structure.
Memory and latency shrink relative to token-based attention at fixed hardware, which makes longer lookback windows and longer forecast horizons more affordable in practice.

Reading between the lines

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

The paper only evaluates forecasting; the same score-network substitution could plausibly extend to other Transformer uses where the score matrix is smooth and the token count is large, such as long-document processing or multivariate anomaly detection.
The linear-complexity claim is asymptotic: because $D$ and $k$ are treated as constants, the practical linear scaling only appears once $L$ and $H$ are much larger than $D^2$, which is not guaranteed in the tested settings.
A direct mechanistic test would be to train CASA on synthetic series with known but weak inter-variate correlations; if the accuracy gain over standard attention disappears, the cross-dimensional score mechanism is the active ingredient.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Desk Editor's Note

read the letter

Short version: the paper is worth engaging, but not for the reason it gives. The CASA module is a channel-gating mechanism; Eq. (8) is an element-wise product, so it cannot reproduce the row-mixing of softmax(QK^T)V. The theoretical story in Section 3.4—that the CNN autoencoder approximates attention scores—does not survive contact with the math. What remains is a solid empirical study of a cheap, modular attention replacement, with extensive results across three tokenization schemes.

The genuinely new piece is the score network: a 1D CNN autoencoder that produces a per-token score map, combined with V element-wise. The integration with point-, patch-, and channel-wise tokenization is a useful contribution, and the experiments are broad: eight datasets, four horizons, 64 metrics, and a 54/64 best-score count. The code is public. The two propositions are correct but near-tautological—they only say affine maps act row-wise—so they add little.

The soft spots are in the framing and the reporting. First, the attention approximation claim is wrong. With Z_i in R^{N x D}, Score(Z_i) must be in R^{N x D}, and the output row i is v_i times a scalar. Convolution lets the scalar for row i depend on other rows, but no information from other variates is added to the output. That is gating, not attention. It may be a good gate, but the paper should say so. Second, the abstract says 87.5% first-place metrics; the appendix and the table say 84.4% (54/64). That is a concrete error. Third, the complexity analysis hides D^2. They call it constant and ignore it, but at D=512 and L*H ~ 69k, D^2 ~ 262k is four times larger. The linear-in-L,H claim is asymptotic, not true in the tested regime. The memory table also does not support the 77.7% headline: on Traffic, CASA is 1684 MB vs SOFTS 1720 MB—about 2%. Finally, no seeds or error bars, and baseline numbers are not tied to their sources, so the SOTA claim is not yet fully verifiable.

For whom: anyone working on efficient multivariate forecasting or attention alternatives. The module is simple enough to drop into an existing model and test. I would like to see a revision that fixes the framing and the reporting; a serious referee can sort that out. I would accept it for review, with the expectation that the authors either drop the approximation claim or justify it, correct the abstract, and report variance.

Referee Report

4 major / 5 minor

Summary. The paper proposes CASA, a module that replaces the self-attention mechanism in Transformer encoders with a 1D CNN autoencoder that computes a per-token score map, applied element-wise to the value matrix. The authors claim that CASA approximates QK^T/sqrt(d_k) at linear complexity, reduces memory usage by up to 77.7%, accelerates inference by 44.0%, and achieves state-of-the-art results on eight multivariate long-term time-series forecasting benchmarks, ranking first in 54 of 64 metrics. The paper also reports experiments where CASA replaces attention in vanilla Transformer, PatchTST, and iTransformer, with improvements in 40 of 42 metrics, a robustness study across input lengths, and a correlation-matrix analysis on the Weather dataset.

Significance. If the theoretical framing is corrected, CASA is a potentially useful empirical contribution: a simple, model-agnostic CNN gating module with linear complexity that improves accuracy over several strong baselines, with code available and a fairly extensive evaluation across eight datasets. The formal propositions are correct but elementary; the main novelty is the module and its empirical validation, not the formal analysis. The paper would be strengthened by honest framing as a gating mechanism, by uncertainty quantification, and by consistent reporting of the abstract's claims with the full results.

major comments (4)

[Section 3.4, Eq. (8)] The paper claims that the CNN autoencoder 'approximates QK^T/sqrt(d_k)', but the implemented operation does not have the form of an attention-score matrix product. In Eq. (8), Attention(Z_{i+1}) = softmax(Score(Z_i)) ⊛ V_{i+1}; with V_{i+1} ∈ R^{N×D} and ⊛ denoting element-wise product, Score(Z_i) is a per-token vector in R^{N×D}, not an N×N pairwise matrix. Standard self-attention in Eq. (4) produces each output row as a weighted sum of all value rows, whereas Eq. (8) only rescales each value row by a scalar gate. The convolution over variates makes the gate for row i depend on other rows, but it never aggregates or mixes values across variates. Therefore the claim that CASA approximates QK^T/sqrt(d_k), and the related narrative about repairing missing cross-variate information in attention scores, is not supported by the equations. Please reframe CASA as a CNN gating mechanism and re-derive the motivation accordingly.
[Abstract vs. Appendix B] The abstract states 'ranking first in 87.5% of evaluated metrics', but Appendix B and Table 6 report that CASA achieves the best performance in 54 out of 64 metrics, which is 84.4%, not 87.5%. The abstract should be corrected to match the full results.
[Section 3.4, Complexity Analysis and Appendix D] The complexity analysis concludes that the method scales linearly in N, L, and H by treating the hidden dimension D and kernel size k as constants, but in the experimental regime D≈512, and k is a small integer, so D^2 and kD^2 terms are comparable to or larger than LD and DH for L=96 and H=720 (LD≈4.9e4, D^2≈2.6e5). The linear scaling is therefore asymptotic, not empirically demonstrated at the tested operating points. In addition, Appendix D states that the Traffic efficiency experiment uses L=512 and H=96, while the forecasting results use L=96; the memory and inference numbers in Table 1 and Figure 6 thus come from a different operating point than the accuracy experiments. Please specify the exact baseline and configuration for the claimed 77.7% memory reduction; Table 1 shows CASA at 1684 MB versus SOFTS at 1720 MB, which is only about a 2% saving.
[Tables 2, 3, and 6] All results are reported without standard deviations, confidence intervals, or the number of seeds. Many of the reported differences are small (for example, ETTm2 average MSE 0.276 versus 0.287 for SOFTS, and ECL average MSE 0.168 versus 0.174 for SOFTS), so a claim of state-of-the-art performance across 64 metrics needs evidence that the improvements are not due to training noise. Please add variance estimates at least for the main results table.

minor comments (5)

[Section 4, Baselines] The word 'comtemporarlly' should be corrected to 'contemporary'.
[Abstract] The word 'model-agnosticically' should be corrected to 'model-agnostically'.
[Section 3.4, Eqs. (8)-(9)] The symbol f is used both for the affine map defining V_{i+1} and, implicitly, for the score network; please use distinct notation for these two functions.
[Table 6] Table 6 reports results for TimesNet, Stationary, and SCINet, which are not listed among the baselines in Section 4; please add the appropriate references or explain the provenance of those numbers.
[Section 4.4, Table 4] The text says the comparison uses 'MSE between their probability density functions (PDFs)', but the table header only says 'MSE'; please clarify the exact computation and metric name in the table.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: CASA's empirical results are benchmarked externally and its complexity analysis is an operation count; the QK^T approximation claim is an unsupported assertion, not a circular reduction.

full rationale

The paper's central derived claims are the complexity/memory efficiency and the forecasting accuracy of CASA. The complexity statement in Section 3.4 is a straightforward operation count: score network O(N k D^2), embedding O(N L D), predictor O(N D H), with total O(N L + N LD + N D^2 + N kD^2 + N DH), and the linear-in-L,H,N claim follows once D and k are treated as constants. Memory and inference figures in Table 1 and Figure 6 are empirical measurements against independent baselines, and the accuracy results in Tables 2 and 6 are MSE/MAE comparisons on eight public datasets. No fitted parameter is later renamed as a prediction, and no load-bearing self-citation appears in the reference list. The statement that CASA 'approximates QK^T/sqrt(d_k)' using a CNN autoencoder is asserted as design motivation rather than derived, and Eq. (8)'s element-wise softmax(Score) ⊛ V may indeed fail to reproduce the row-mixing of softmax(QK^T)V; that is a correctness and interpretability concern, not a circularity concern, because the architecture's output is not constructed to equal the benchmark results. The paper also contains reporting inconsistencies (e.g., Appendix D lists L=512, H=96 while the main setup uses L=96; the 77.7% memory reduction is not explicitly tied to a table row), but these are presentation issues rather than circular reductions. Overall, the central empirical and efficiency claims are self-contained and externally checkable, so the circularity score is 0.

Assumptions & free parameters 2 free parameters · 3 assumptions · 0 invented entities

The model introduces a trainable CNN autoencoder score network, which is an architectural component rather than a separate postulated entity. The free parameters (D, k) are standard architectural choices, not fitted numbers used to force a result. The domain assumptions are stated implicitly through the method design and benchmark selection.

free parameters (2)

Hidden dimension D = not reported (commonly 512)
The model capacity and the dominant O(N D^2) term in the complexity analysis depend on D; the paper does not report D for each dataset.
CNN kernel size k = not reported
The kernel size controls the receptive field over channels; the paper mentions it in Appendix D but does not state the value used.

assumptions (3)

standard math Softmax normalization and element-wise multiplication produce a valid attention-like weighting.
Used in Eq. 8 to define CASA attention; this is a standard operation.
domain assumption Convolution along the channel dimension captures cross-variate dependencies in multivariate time series.
This is the core motivation in Section 3.4; the paper provides empirical correlation-matrix evidence but no formal guarantee.
domain assumption The eight benchmark datasets are representative of multivariate LTSF problems.
The empirical evaluation and SOTA claim in Section 4 rely on this.

how reviews work

0 comments

Cite this review

Pith. "Pith review of CASA: CNN Autoencoder-based Score Attention for Efficient Multivariate Long-term Time-series Forecasting." pith.science (2026). https://pith.science/paper/URJD6M54

@misc{pith2026250502011,
  author       = {Pith},
  title        = {Pith review of: CASA: CNN Autoencoder-based Score Attention for Efficient Multivariate Long-term Time-series Forecasting},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/URJD6M54}},
  note         = {Machine review of arXiv:2505.02011}
}

read the original abstract

Multivariate long-term time series forecasting is critical for applications such as weather prediction, and traffic analysis. In addition, the implementation of Transformer variants has improved prediction accuracy. Following these variants, different input data process approaches also enhanced the field, such as tokenization techniques including point-wise, channel-wise, and patch-wise tokenization. However, previous studies still have limitations in time complexity, computational resources, and cross-dimensional interactions. To address these limitations, we introduce a novel CNN Autoencoder-based Score Attention mechanism (CASA), which can be introduced in diverse Transformers model-agnosticically by reducing memory and leading to improvement in model performance. Experiments on eight real-world datasets validate that CASA decreases computational resources by up to 77.7%, accelerates inference by 44.0%, and achieves state-of-the-art performance, ranking first in 87.5% of evaluated metrics.

Figures

Figures reproduced from arXiv: 2505.02011 by the authors.

**Figure 2.** (a) point-wise token (b) patch-wise token (c) channel-wise token [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗

**Figure 3.** (a) Conventional Self-Attention. (b) Overall architecture of our CASA block. The time-series data is embedded using channel-wise [PITH_FULL_IMAGE:figures/full_fig_p003_3.png] view at source ↗

Figures from the paper (11 more)

**Figure 4.** Figure 4: Results of prediction of Ours and baseline models on Weather dataset. (a) CASA (b) iTransformer (c) SOFTS (d) PatchTST [PITH_FULL_IMAGE:figures/full_fig_p005_4.png]

**Figure 5.** Figure 5: Experimental results on the ETTm1, Electricity, and Traffic datasets (with 7, 321, and 862 variates, respectively). Our CASA [PITH_FULL_IMAGE:figures/full_fig_p006_5.png]

**Figure 6.** Figure 6: (a) Memory usage scaling with the number of tokens, demonstrating CASA’s linear growth and reduced memory consumption [PITH_FULL_IMAGE:figures/full_fig_p006_6.png]

**Figure 7.** Figure 7: (a) Correlation matrix among the Ground Truth variates, [PITH_FULL_IMAGE:figures/full_fig_p007_7.png]

**Figure 8.** Figure 8: Results of prediction of Ours and baseline models on Etth1 dataset [PITH_FULL_IMAGE:figures/full_fig_p010_8.png]

**Figure 9.** Figure 9: Results of prediction of Ours and baseline models on Etth2 dataset [PITH_FULL_IMAGE:figures/full_fig_p010_9.png]

**Figure 10.** Figure 10: Results of prediction of Ours and baseline models on Ettm1 dataset [PITH_FULL_IMAGE:figures/full_fig_p010_10.png]

**Figure 11.** Figure 11: Results of prediction of Ours and baseline models on Ettm2 dataset [PITH_FULL_IMAGE:figures/full_fig_p010_11.png]

**Figure 12.** Figure 12: Results of prediction of Ours and baseline models on Traffic dataset [PITH_FULL_IMAGE:figures/full_fig_p011_12.png]

**Figure 13.** Figure 13: Results of prediction of Ours and baseline models on Solar dataset [PITH_FULL_IMAGE:figures/full_fig_p011_13.png]

**Figure 14.** Figure 14: Results of prediction of Ours and baseline models on Electricity dataset [PITH_FULL_IMAGE:figures/full_fig_p011_14.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

25 extracted references · 17 canonical work pages

[1]

Representation learning: A review and new perspectives

[Bengio et al., 2013] Yoshua Bengio, Aaron Courville, and Pascal Vincent. Representation learning: A review and new perspectives. IEEE transactions on pattern analysis and machine intelligence, 35(8):1798–1828,

work page 2013
[4]

Synthetic data in health care: A narrative review

[Gonzales et al., 2023] Aldren Gonzales, Guruprabha Gu- ruswamy, and Scott R Smith. Synthetic data in health care: A narrative review. PLOS Digital Health, 2(1):e0000082,

work page 2023
[5]

Softs: Efficient multivariate time series forecasting with series-core fusion

[Han et al., 2024] Lu Han, Xu-Yang Chen, Han-Jia Ye, and De-Chuan Zhan. Softs: Efficient multivariate time series forecasting with series-core fusion. In the twelfth interna- tional conference on learning representations,

work page 2024
[7]

arXiv preprint arXiv:2401.01629,

arXiv
[8]

Spatio-temporal self-supervised learning for traffic flow prediction

[Ji et al., 2023] Jiahao Ji, Jingyuan Wang, Chao Huang, Junjie Wu, Boren Xu, Zhenhe Wu, Junbo Zhang, and Yu Zheng. Spatio-temporal self-supervised learning for traffic flow prediction. In Proceedings of the AAAI con- ference on artificial intelligence, volume 37, pages 4356– 4364,

work page 2023
[9]

Modeling long-and short-term temporal patterns with deep neural networks

[Lai et al., 2018] Guokun Lai, Wei-Cheng Chang, Yiming Yang, and Hanxiao Liu. Modeling long-and short-term temporal patterns with deep neural networks. In The 41st international ACM SIGIR conference on research & devel- opment in information retrieval, pages 95–104,

2018
[11]

A time series is worth 64 words: Long-term forecasting with transformers

[Nie et al., 2022] Yuqi Nie, Nam H Nguyen, Phanwadee Sinthong, and Jayant Kalagnanam. A time series is worth 64 words: Long-term forecasting with transformers. In the tenth international conference on learning representa- tions,

work page 2022
[12]

Attention is all you need

[Vaswani, 2017] A Vaswani. Attention is all you need. Ad- vances in Neural Information Processing Systems,

2017

Show all 25 references

[16]

Re- vitalizing multivariate time series forecasting: Learnable decomposition with inter-series dependencies and intra- series variations modeling

[Yu et al., 2024] Guoqi Yu, Jing Zou, Xiaowei Hu, Angel- ica I Aviles-Rivero, Jing Qin, and Shujun Wang. Re- vitalizing multivariate time series forecasting: Learnable decomposition with inter-series dependencies and intra- series variations modeling. International conference ...

2024
[17]

Are transformers effective for time series fore- casting? In Proceedings of the AAAI conference on artifi- cial intelligence, volume 37, pages 11121–11128,

[Zeng et al., 2023] Ailing Zeng, Muxi Chen, Lei Zhang, and Qiang Xu. Are transformers effective for time series fore- casting? In Proceedings of the AAAI conference on artifi- cial intelligence, volume 37, pages 11121–11128,

2023
[18]

Crossformer: Transformer utilizing cross-dimension de- pendency for multivariate time series forecasting

[Zhang and Yan, 2023] Yunhao Zhang and Junchi Yan. Crossformer: Transformer utilizing cross-dimension de- pendency for multivariate time series forecasting. In the eleventh international conference on learning representa- tions,

2023
[19]

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

[Zhou et al., 2021] Haoyi Zhou, Shanghang Zhang, Jieqi Peng, Shuai Zhang, Jianxin Li, Hui Xiong, and Wan- cai Zhang. Informer: Beyond efficient transformer for long sequence time-series forecasting. In Proceedings of the AAAI conference on artificial intelligence, volume 35, p...

2021
[20]

Fedformer: Fre- quency enhanced decomposed transformer for long-term series forecasting

[Zhou et al., 2022] Tian Zhou, Ziqing Ma, Qingsong Wen, Xue Wang, Liang Sun, and Rong Jin. Fedformer: Fre- quency enhanced decomposed transformer for long-term series forecasting. In International conference on machine learning, pages 27268–27286. PMLR,

2022
[21]

Appendix A Dataset A.1 Details of Each Dataset Data preprocessing is largely based on previous work SOFTS. We provide the dataset descriptions: • ETT (Electricity Transformer Temperature) : This dataset includes two hourly (ETTh) and two 15-minute (ETTm) variations, containing...

2016
[23]

• Electricity: It consists of hourly electricity usage data for 321 consumers, spanning from 2012 to

2012
[25]

Dimensions denotes the number of vari- ate in each dataset

Dataset Timesteps Sample Frequency Dimension Weather 52,696 10 min 21 Electricity 26,304 1 hour 321 Traffic 17,544 1 hour 862 Solar 52,560 10 min 137 ETTh1 17,420 1 hour 7 ETTh2 17,420 1 hour 7 ETTm1 69,680 15 min 7 ETTm2 69,680 15 min 7 Table 5: Characteristics of the dataset...

2024
[2013]

Long-term forecasting with tide: Time-series dense encoder

[Das et al., 2023] Abhimanyu Das, Weihao Kong, Andrew Leach, Shaan Mathur, Rajat Sen, and Rose Yu. Long-term forecasting with tide: Time-series dense encoder. arXiv preprint arXiv:2304.08424,

2023 arXiv
[2014]

• Solar-Energy: It captures solar power generation data from 137 photovoltaic plants in 2006, sampled every 10 minutes

• Weather: it contains 21 weather-related variables, such as temperature and humidity, recorded every 10 minutes throughout 2020 in Germany. • Solar-Energy: It captures solar power generation data from 137 photovoltaic plants in 2006, sampled every 10 minutes. Further details ...

2020
[2016]

Autoformer: Decomposition transform- ers with auto-correlation for long-term series forecast- ing

[Wu et al., 2021] Haixu Wu, Jiehui Xu, Jianmin Wang, and Mingsheng Long. Autoformer: Decomposition transform- ers with auto-correlation for long-term series forecast- ing. Advances in neural information processing systems , 34:22419–22430,

2021
[2017]

Deep kernel learn- ing

[Wilson et al., 2016] Andrew Gordon Wilson, Zhiting Hu, Ruslan Salakhutdinov, and Eric P Xing. Deep kernel learn- ing. In Artificial intelligence and statistics , pages 370–

2016
[2018]

• Traffic: This dataset represents hourly road occupancy rates gathered by sensors monitoring San Francisco free- ways from 2015 to

2015
[2021]

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

[Wu et al., 2023] Haixu Wu, Tengge Hu, Yong Liu, Hang Zhou, Jianmin Wang, and Mingsheng Long. Timesnet: Temporal 2d-variation modeling for general time series analysis. In the eleventh international conference on learning representations,

2023
[2022]

itransformer: Inverted transformers are effective for time series forecasting

[Liu et al., 2023] Yong Liu, Tengge Hu, Haoran Zhang, Haixu Wu, Shiyu Wang, Lintao Ma, and Mingsheng Long. itransformer: Inverted transformers are effective for time series forecasting. the eleventh international conference on learning representations,

2023
[2023]

Tsmixer: Lightweight mlp-mixer model for multivariate time series forecasting

[Ekambaram et al., 2023] Vijay Ekambaram, Arindam Jati, Nam Nguyen, Phanwadee Sinthong, and Jayant Kalagnanam. Tsmixer: Lightweight mlp-mixer model for multivariate time series forecasting. In Proceedings of the 29th ACM SIGKDD Conference on Knowledge Discovery and Data Mining...

2023
[2024]

Synthetic data in ai: Chal- lenges, applications, and ethical implications

[Hao et al., 2024] S Hao, W Han, T Jiang, Y Li, H Wu, C Zhong, Z Zhou, and H Tang. Synthetic data in ai: Chal- lenges, applications, and ethical implications. arxiv

2024

Pith tools

Reviewed August 16, 2026 · model on record in the stance chip above.

[1] [1]

Representation learning: A review and new perspectives

[Bengio et al., 2013] Yoshua Bengio, Aaron Courville, and Pascal Vincent. Representation learning: A review and new perspectives. IEEE transactions on pattern analysis and machine intelligence, 35(8):1798–1828,

work page 2013

[2] [4]

Synthetic data in health care: A narrative review

[Gonzales et al., 2023] Aldren Gonzales, Guruprabha Gu- ruswamy, and Scott R Smith. Synthetic data in health care: A narrative review. PLOS Digital Health, 2(1):e0000082,

work page 2023

[3] [5]

Softs: Efficient multivariate time series forecasting with series-core fusion

[Han et al., 2024] Lu Han, Xu-Yang Chen, Han-Jia Ye, and De-Chuan Zhan. Softs: Efficient multivariate time series forecasting with series-core fusion. In the twelfth interna- tional conference on learning representations,

work page 2024

[4] [7]

arXiv preprint arXiv:2401.01629,

arXiv

[5] [8]

Spatio-temporal self-supervised learning for traffic flow prediction

[Ji et al., 2023] Jiahao Ji, Jingyuan Wang, Chao Huang, Junjie Wu, Boren Xu, Zhenhe Wu, Junbo Zhang, and Yu Zheng. Spatio-temporal self-supervised learning for traffic flow prediction. In Proceedings of the AAAI con- ference on artificial intelligence, volume 37, pages 4356– 4364,

work page 2023

[6] [9]

Modeling long-and short-term temporal patterns with deep neural networks

[Lai et al., 2018] Guokun Lai, Wei-Cheng Chang, Yiming Yang, and Hanxiao Liu. Modeling long-and short-term temporal patterns with deep neural networks. In The 41st international ACM SIGIR conference on research & devel- opment in information retrieval, pages 95–104,

2018

[7] [11]

A time series is worth 64 words: Long-term forecasting with transformers

[Nie et al., 2022] Yuqi Nie, Nam H Nguyen, Phanwadee Sinthong, and Jayant Kalagnanam. A time series is worth 64 words: Long-term forecasting with transformers. In the tenth international conference on learning representa- tions,

work page 2022

[8] [12]

Attention is all you need

[Vaswani, 2017] A Vaswani. Attention is all you need. Ad- vances in Neural Information Processing Systems,

2017