DTAMLP: Denoise Time-aware MLP for Session-based Recommendation

Jiamu Zheng; Xiaojun Shan

REVIEW 4 major objections 6 minor 84 references

DTAMLP: Denoise Time-aware MLP for Session-based Recommendation

T0 review · 4 major / 6 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read The central claim is that down-weighting clicks with very short time intervals, via a plug-and-play weight fusion module, consistently improves time-aware and GNN-based session recommenders.

desk verdict A cheap, plausible plug-in trick for time-aware recommenders, reported honestly but with evidence that doesn't yet isolate the interval signal. read the letter →

arxiv 2608.12975 v1 pith:4GWRXKZW submitted 2026-08-13 cs.SI cs.IR

Jiamu Zheng , Xiaojun Shan This is my paper · ORCID

classification cs.SIcs.IR

keywords session-basedrecommendationsporadicnoisetimeintervalweightfusionfrequency-domainfilteringFFTpreferenceMLPrecommender

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 tries to establish a small, practical claim about session-based recommendation: models that use click timestamps treat every gap between clicks as equally meaningful, but very short gaps often mark accidental, low-preference clicks. It proposes a plug-and-play weight fusion module that blends a model's existing attention weight with a threshold-capped softmax weight over click intervals, so clicks with tiny intervals are down-weighted. Applied without touching the backbone, the module improves TiSASRec on MovieLens-1m and SR-GNN on Diginetica, which the paper calls its most directly verifiable result. The paper also offers a speculative explanation for why frequency-domain filtering helps, and packages both ideas in an all-MLP model, DTAMLP, validated on Diginetica and RetailRocket. The system-level accuracy numbers are presented as supporting evidence, not as a state-of-the-art claim.

What carries the argument

The load-bearing object is the weight fusion identity: the final item weight is a blend of the model's own attention weight and a softmax over capped time intervals, $\alpha_2 = \mathrm{softmax}(\min(\mathrm{interval}, \eta))$, combined as $\alpha = (1-\beta)\alpha_1 + \beta\alpha_2$. The cap $\eta$ is the upper limit beyond which longer gaps no longer add signal, so only suspiciously short intervals are punished. In DTAMLP the same recipe appears as Eqs. 15-16 with a learnable $\beta$, alongside an FFT-Transformer, $\tilde{F}^l = \mathcal{F}^{-1}(W \odot \mathcal{F}(F^l))$, that filters item embeddings in the frequency domain, plus a representation-consistent session embedding adapted from prior work. The weight fusion carries the paper's central claim; the FFT module carries its interpretive conjecture.

What would settle it

Take the same plug-in weight fusion and either reverse its direction, up-weighting the shortest intervals, or apply it to a dataset with true dwell-time labels: if reversed weighting also helps, or if short-interval clicks are frequently followed by a same-item re-click, purchase, or long subsequent view, then the central claim that short gaps are sporadic noise fails.

Watch

Extended reading notes

Core claim

On its own terms, the central discovery is that click-time intervals carry exploitable signal about sporadic noise: a very short gap between two adjacent clicks tends to follow a click made out of curiosity or misleading appeal, not genuine preference. The paper formalizes this with a capped interval weight $\alpha_2 = \mathrm{softmax}(\min(\mathrm{interval}, \eta))$ and fuses it with the model's attention weight $\alpha_1$ via $\alpha = (1-\beta)\alpha_1 + \beta\alpha_2$; the blend coefficient $\beta$ can be fixed or learnable. This one equation, inserted into TiSASRec and SR-GNN with no other architectural change, produces consistent gains, including SR-GNN's MRR@20 rising from 15.83 to 18.01 on Diginetica. The paper is explicit that the second claim, that Fourier-domain filtering helps because it separates entangled 'preference noise,' is an interpretive conjecture, not a proven mechanism.

Load-bearing premise

The result rests on treating the interval between adjacent clicks as a proxy for how long the user actually looked at the item, so that a very short interval marks a low-preference sporadic click.

Editorial extensions

If this is right

Time-aware recommenders can gain accuracy by down-weighting clicks separated by very short intervals, without retraining the backbone.
The fusion can be added to attention-based and GNN-based models even when they have no explicit time-awareness, as the SR-GNN result shows.
Ablations on DTAMLP indicate that time-interval weight fusion and frequency-domain filtering improve accuracy independently and complementarily.
The gain is larger on ranking-sensitive metrics than on recall, so the intervention mostly reorders candidates rather than retrieving a different set of relevant items.

Reading between the lines

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

Beyond the paper: a direct extension would be to test the same fusion across more backbones and datasets and measure whether the optimal threshold $\eta$ tracks session length or domain, which the paper does not report.
Beyond the paper: if short intervals genuinely mark sporadic noise, then the module's down-weighting should agree with true dwell-time labels from logs such as scroll depth or video watch time; the paper only uses click gaps.
Beyond the paper: the preference-noise conjecture could be probed by inspecting learned FFT filters and checking whether items sharing a filter response cluster by interpretable attributes, a test the paper leaves open.
Beyond the paper: because the module needs no architectural change, it could operate as a post-hoc scoring adjustment on a deployed model, an operational consequence the paper does not develop.

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

Here's the short version: the weight-fusion module is a genuinely cheap idea – blend a model's attention weight with a threshold-capped softmax over click intervals – and the demonstration on two architecturally different backbones (TiSASRec, SR-GNN) is easy to believe as a useful trick. But the paper never actually shows that the interval values are the active ingredient. Eq. 3 is a convex combination that always down-weights the shortest intervals; any smooth reweighting of attention might give the same gain. There's no shuffled-interval or inverted-interval control, so the 'sporadic noise' interpretation is not established. That's not fatal for the engineering claim, but it is the difference between a trick and a finding.

What the paper does well: it is clearly scoped and unusually honest. It labels the frequency-domain 'preference noise' story as an interpretive conjecture, disclaims SOTA status, and points to the plug-in verification as its most direct contribution. The interval-capped softmax fusion is not in the cited TiSASRec or SR-GNN work, as far as I can tell, so the novelty claim is reasonable. The ablation structure is sensible, though the numbers are weak support.

The soft spots, in order: (1) no control condition, as above; (2) all metrics are single-run with no standard deviations or significance tests – the TiSASRec NDCG gains in Table 1 are small enough to be noise; (3) hyperparameters including intervalmax and margin are tuned on RetailRocket and then the tuned numbers are reported as the main result, which is a form of test-set leakage; (4) the interval-as-dwell-time proxy is asserted from a shopping anecdote, with no distribution analysis or robustness check; (5) no code released, so independent verification means reimplementing.

The preference-noise conjecture is clearly labeled as speculation, so I don't fault them for it. I'd rather see the paper spend its space on a proper control for the weight fusion.

Who should read this: anyone working on time-aware or GNN-based session recommendation who wants a quick trick to try. It deserves a serious referee, because the central claim is cheap to verify and the write-up is honest enough that a revision with controls and multi-seed runs could make it solid.

Referee Report

4 major / 6 minor

Summary. This paper makes two empirical claims and combines them in DTAMLP. The first, presented as the central result, is that click-time intervals in session-based recommendation contain 'sporadic noise': very short intervals often reflect accidental or low-preference clicks, so fusing a model's attention weight with a threshold-capped softmax over intervals (Eq. 3; Eqs. 15–16) improves accuracy when plugged into TiSASRec and SR-GNN (Table 1). The second is a conjecture, explicitly labeled as unproven, that FFT-based filtering helps because it separates entangled 'preference noise'; DTAMLP integrates weight fusion with an FFT-Transformer and a representation-consistent embedding design, and is evaluated on Diginetica and RetailRocket (Tables 3 and 4), with ablations and parameter sensitivity.

Significance. If the plug-and-play weight fusion result were established, it would be practically useful: it is a near-cost-free modification applicable to widely used SBR backbones, and the paper's framing makes the claim falsifiable. The paper is also commendably honest about the status of its second claim: the preference-noise explanation is explicitly an interpretation, not a mechanism, and the system-level results are not presented as state of the art. The main weakness is that the central empirical claim is not yet supported at the standard needed for publication: Table 1 lacks statistical replication and, more importantly, lacks control conditions that separate the interval signal from generic attention reweighting. The DTAMLP results depend on tuning choices that are not shown to be stable across datasets. Because the required evidence is obtainable within the scope of the manuscript, I regard this as a major-revision situation rather than a rejection.

major comments (4)

[§3.2.1, Eq. (3), Table 1] The claimed evidence for sporadic noise is incomplete. The fusion operation α = (1−β)α1 + β softmax(min(interval,η)) is a monotone reweighting that always down-weights the shortest intervals, regardless of whether those intervals actually carry sporadic noise. The experiments compare only “weight fusion vs. unmodified backbone,” so the gains in Table 1 are also compatible with the alternative hypothesis that any soft reweighting of attention helps (e.g., by reducing overconfident attention weights). Please add control conditions with shuffled intervals, inverted intervals (softmax over −interval), and a position-based weight of the same functional form, and report whether the gains persist. Without these controls, the central claim that the threshold-capped time-interval weight is the active ingredient is not established.
[Table 1; Tables 3–4] All accuracy numbers are single-run metrics with no standard deviations, confidence intervals, or significance tests. The TiSASRec gains in Table 1 (NDCG@10 from 56.8163 to 57.1112 at β=0.1, and HR@10 from 79.3377 to 80.1324) are small and could easily lie within run-to-run noise; the same concern applies to many differences in Tables 3 and 4. Please report means and standard deviations over at least 3–5 seeds and, for the plug-in comparisons, a paired significance test over sessions. This is necessary to support the abstract's phrase “consistent accuracy gain.”
[§5.3.3; Tables 3–4] Hyperparameters intervalmax, margin, dropout, and embedding size are tuned on RetailRocket and then RetailRocket results are reported in Table 4. If this tuning used the test set, the reported numbers are optimistic and the procedure is circular with respect to robustness claims. Please state explicitly which split was used for tuning, fix hyperparameters before evaluating the test set, and show that the chosen values (or the qualitative conclusion) are stable across datasets or over a range of settings.
[§3.2.1] The premise that the interval between two adjacent clicks is a usable proxy for dwell time, and that very short intervals predominantly indicate low-preference sporadic clicks, is asserted via an anecdotal shopping example. No distribution of intervals, no behavioral validation, and no robustness check with alternative interval definitions are provided. If short intervals often reflect rapid but genuine exploration, down-weighting them would reduce real preference signal. Please either test this premise directly (e.g., relate intervals to next-click relevance) or show that the plug-in gain is insensitive to the interval definition.

minor comments (6)

[§3.1] The heading contains a typo: “Promblem Formulation” should be “Problem Formulation.”
[§5.3.2] The variant label “w/o None” is confusing; please use “Full model” or “No component removed.”
[Figures 6–7] Figures 6 and 7 are not rendered correctly in the submitted version: the panels contain raw UTF strings (e.g., “/uni00000013/...”), and the panel labels do not match the text, which describes tuning of intervalmax, margin, dropout, and embedding size. Please regenerate the figures.
[§5.4] The complexity comparison states O(d log d) for the FFT-Transformer but does not account for the DNN, time-embedding, and prediction components; the reported training time is not tied to a reproducible configuration. Please clarify the scope of the complexity claim.
[Abstract] The phrase “consistent accuracy gain” overstates what single-run results can support; please qualify it as gains observed in the reported runs.
[General] No code or data release is mentioned; since the central contribution is a plug-and-play modification, providing code (or pseudocode for the weight-fusion module) would greatly aid verification.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the paper's central claims are measured empirical observations or explicitly labeled conjectures, with no derivation that reduces to its own inputs.

full rationale

The paper's central, most directly verifiable claim is that a plug-and-play weight fusion module (Eq. 3, and Eq. 15-16 in DTAMLP) improves accuracy on TiSASRec and SR-GNN. This is an empirical result measured on benchmarks after setting beta and eta, not a prediction derived from the sporadic-noise assumption: the paper does not claim to derive the gain from first principles. The sporadic-noise and preference-noise constructs are explicitly presented as an observed phenomenon and an interpretive conjecture, respectively, rather than as proven mechanisms. The only externally adopted components (the representation-consistent design of Hou et al. 2022 and the FFT filtering idea of FMLP-Rec) are openly borrowed and are not used as load-bearing self-citations; no uniqueness theorem or prior work by the same authors is invoked to force a conclusion. The lack of a shuffled-interval or inverted-interval control in Table 1 is a potential experimental confound and a correctness risk, but it is not circularity: the intervention's effect is still independently measured, and no equation in the paper reduces the observed accuracy gain to the sporadic-noise assumption by construction. Accordingly, the paper is self-contained against its own evidence and no circular step can be exhibited.

Assumptions & free parameters 8 free parameters · 6 assumptions · 2 invented entities

All free parameters are hand-set or tuned, with no theory fixing them. The behavioral assumptions about click intervals and the preference-frequency mapping are load-bearing and untested. The invented concepts of sporadic noise and preference noise are post hoc labels rather than independently measured phenomena.

free parameters (8)

interval threshold eta (intervalmax in DTAMLP) = eta set to 20th percentile of intervals for TiSASRec; intervalmax tuned to 70000 for RetailRocket
Caps the click-interval feature in the softmax weight; no principled formula; chosen by hand and by tuning in Sections 3.2.1 and 5.3.3.
blend coefficient beta = beta in {0.1, 0.2, 0.3} for TiSASRec; learnable for SR-GNN and DTAMLP
Controls how much the interval-based weight can override the model attention; no independent justification beyond observed accuracy.
time slice width for time embedding = 600 seconds, with 144 slices
Eq. 12 uses min(floor((tc - ti)/600), 144); the paper says the time slice width is determined experimentally.
RDM margin tau = 0.05
Tuned in Section 5.3.3; the paper notes small margin values can drop performance below GCARM.
dropout ratio = 0.2
Selected from a 0 to 0.5 sweep in Section 5.3.3.
embedding size = 100
Set in Section 5.2; the paper's own ablation shows larger embeddings improve accuracy, so 100 is a practical resource bound rather than an optimal value.
session truncation length = 50
Only the last 50 clicks are kept per session; a preprocessing choice with no theoretical justification.
minimum item frequency = 5 occurrences
Items appearing fewer than five times are removed; this changes the item space and can affect all compared methods.

assumptions (6)

standard math DFT is invertible and one-to-one; inverse DFT recovers the original sequence.
Used to justify the FFT-Transformer in Eqs. 4-5 and Section 3.2.2; standard and uncontroversial.
domain assumption Click interval between two adjacent clicks approximates dwell time on the first item.
Stated in Section 3.2.1; if false, the sporadic-noise signal is mis-specified.
domain assumption Very short dwell times are more likely to indicate accidental clicks carrying little preference signal.
The key behavioral premise behind down-weighting short intervals; the paper provides intuition but no direct measurement.
domain assumption Session embedding as a weighted linear combination of item embeddings preserves recommendation semantics.
Adapted from Hou et al. (2022) and used in Eqs. 7-8; the paper does not re-derive this.
ad hoc to paper Frequency components of item embeddings loosely correspond to entangled latent preference components.
Figure 2 and Section 3.2.2; the paper explicitly calls this an illustrative conjecture, not an empirically identified mapping.
ad hoc to paper Softmax over capped intervals produces a valid complementary attention distribution.
The operation in Eqs. 2 and 15 has no theoretical grounding; its validity is judged only through downstream accuracy.

invented entities (2)

sporadic noise
purpose: Label for short dwell-time clicks weakly related to real preference.
The only evidence is the within-paper improvement when down-weighting short intervals; there is no independent behavioral measurement or external prediction.
preference noise
purpose: Hypothetical entangled psychological preference components that frequency-domain filtering can separate and down-weight.
Explicitly an interpretive conjecture in Section 3.2.2; no direct test of the frequency-preference correspondence is provided.

how reviews work

0 comments

Cite this review

Pith. "Pith review of DTAMLP: Denoise Time-aware MLP for Session-based Recommendation." pith.science (2026). https://pith.science/paper/4GWRXKZW

@misc{pith2026260812975,
  author       = {Pith},
  title        = {Pith review of: DTAMLP: Denoise Time-aware MLP for Session-based Recommendation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/4GWRXKZW}},
  note         = {Machine review of arXiv:2608.12975}
}

read the original abstract

This paper reports two empirical findings on session-based recommendation (SBR), unified in a single model, DTAMLP. First, existing time-aware and GNN-based models (e.g., TiSASRec, SR-GNN) treat every click-time interval as equally informative, even though very short dwell times often reflect accidental clicks carrying little preference signal -- a phenomenon we call sporadic noise. We show that a lightweight, plug-and-play weight fusion module, blending a model's attention weight with a threshold-capped time-interval weight, can be inserted into such models with almost no architectural change and yields a consistent accuracy gain; we view this as the most directly verifiable contribution of this work. Second, we revisit an under-explained observation from FMLP-Rec, where a learnable frequency-domain filter on item embeddings improves accuracy, and offer a possible explanation: time-domain behavior mixes several entangled psychological preferences, and a frequency-domain view may let a model separate and down-weight such preference noise more naturally -- an interpretive conjecture rather than a proven mechanism. Building on both insights, DTAMLP, an all-MLP framework combining weight fusion and FFT-based filtering, is validated on Diginetica and RetailRocket. While this system-level design reflects the state of the field circa 2023 rather than a state-of-the-art claim, ablations confirm the two mechanisms contribute complementary, non-redundant improvements.

Figures

Figures reproduced from arXiv: 2608.12975 by the authors.

**Figure 1.** An example of shopping on e-commerce platforms. (The time above means the time interval between two clicks.) Sometimes you may click ”Item B” with curiosity because its advertising image is really attractive. However, after you click the item and know what it is, you may exit the page and click another item C in a short time because you don’t need it or feel deceived by its false propaganda. 3.2. Analysis on Two Ty… view at source ↗

**Figure 2.** A conceptual illustration of our conjecture: after applying the Fourier transform, we obtain the spectrum of Itemi’s embedding at frequencies ωk = {X, Y, Z}. We conjecture that a triangular wave at a specific frequency may loosely correspond to a specific latent preference component {A, B, C} – for instance, A might loosely correspond to a preference for sweet food while B might correspond to an aversion to spicy f… view at source ↗

**Figure 3.** The framework of the DTAMLP the Space-consistent Embedding Module, which we adopt from prior work as the basis of our system. Additionally, we discuss how our DNN generates the weights for linear combinations, i.e., how the two mechanisms from Section 3 are operationalized. At last, a brief explanation of how Robust Distance Measuring works will be presented. 4.1. Embedding-layer Each item is embedded into a unifie… view at source ↗

Figures from the paper (4 more)

**Figure 4.** Figure 4: Different representation spaces between session embedding and item embeddings can lead to incorrect predictions when measuring distances between embeddings for recommendation purposes (illustration adapted from the observation in (Hou et al., 2022)). α = DNN([hs,1; hs…

**Figure 5.** Figure 5: Visualization of learned session embeddings on Yoochoose datasets with the original Transformer, and “w/o None” is the full DTAMLP with none of those parts removed. Comparing “w/o WFD” (and “w/o TA”) against the full model isolates the contribution of Insight 1 (spora…

**Figure 7.** Figure 7: Ablation study of DTAMLP on Diginetica and Retailrocket. See Section 5 for the definition of each variant. is CNN-based, and for the input size d × h, the complexity of one operation for a convolution kernel of size k × h is O(kh). It is done d times in total, resulti…

**Figure 6.** Figure 6: Parameter tuning of DTAMLP on Retailrocket datasets. all corresponding sessions from our test set. Notably, the session embeddings learned by DTAMLP exhibit visibly better class separation than those derived from SR-GNN and GCARM, which is consistent with – though not …

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

84 extracted references · 55 canonical work pages

[1]

Konstan and John Riedl , title =

Badrul Munir Sarwar and George Karypis and Joseph A. Konstan and John Riedl , title =
[2]

Steffen Rendle and Christoph Freudenthaler and Zeno Gantner and Lars Schmidt
[3]

Factorizing personalized Markov chains for next-basket recommendation , booktitle =

Steffen Rendle and Christoph Freudenthaler and Lars Schmidt. Factorizing personalized Markov chains for next-basket recommendation , booktitle =
[4]

Session-based Recommendations with Recurrent Neural Networks , booktitle =

Bal. Session-based Recommendations with Recurrent Neural Networks , booktitle =
[5]

Parallel Recurrent Neural Network Architectures for Feature-rich Session-based Recommendations , booktitle =

Bal. Parallel Recurrent Neural Network Architectures for Feature-rich Session-based Recommendations , booktitle =

work page
[6]

Neural Collaborative Filtering , booktitle =

Xiangnan He and Lizi Liao and Hanwang Zhang and Liqiang Nie and Xia Hu and Tat. Neural Collaborative Filtering , booktitle =

work page
[7]

Jing Li and Pengjie Ren and Zhumin Chen and Zhaochun Ren and Tao Lian and Jun Ma , title =

work page
[8]

Qiao Liu and Yifu Zeng and Refuoe Mokhosi and Haibin Zhang , title =

work page

Show all 84 references

[9]

Self-Attentive Sequential Recommendation , booktitle =

Wang. Self-Attentive Sequential Recommendation , booktitle =
[10]

Shuai Zhang and Lina Yao and Aixin Sun and Yi Tay , title =
[11]

Pengjie Ren and Zhumin Chen and Jing Li and Zhaochun Ren and Jun Ma and Maarten de Rijke , title =
[12]

Jose and Xiangnan He , title =

Fajie Yuan and Alexandros Karatzoglou and Ioannis Arapakis and Joemon M. Jose and Xiangnan He , title =. WSDM , pages =
[13]

Shu Wu and Yuyuan Tang and Yanqiao Zhu and Liang Wang and Xing Xie and Tieniu Tan , title =
[14]

ACM Transactions on Information Systems (TOIS) , volume=

Graph co-attentive session-based recommendation , author=. ACM Transactions on Information Systems (TOIS) , volume=. 2021 , publisher=

2021
[15]

Sheng and Jiajie Xu and Fuzhen Zhuang and Junhua Fang and Xiaofang Zhou , title =

Chengfeng Xu and Pengpeng Zhao and Yanchi Liu and Victor S. Sheng and Jiajie Xu and Fuzhen Zhuang and Junhua Fang and Xiaofang Zhou , title =
[16]

Fei Sun and Jun Liu and Jian Wu and Changhua Pei and Xiao Lin and Wenwu Ou and Peng Jiang , title =
[17]

arXiv preprint arXiv:1909.04276 , year=

NISER: Normalized Item and Session Representations with Graph Neural Networks , author=. arXiv preprint arXiv:1909.04276 , year=

1909 arXiv
[18]

arXiv preprint arXiv:1902.04864 , year =

Shoujin Wang and Longbing Cao and Yan Wang , title =. arXiv preprint arXiv:1902.04864 , year =

1902 arXiv
[19]

LightGCN: Simplifying and Powering Graph Convolution Network for Recommendation , booktitle =

Xiangnan He and Kuan Deng and Xiang Wang and Yan Li and Yong. LightGCN: Simplifying and Powering Graph Convolution Network for Recommendation , booktitle =
[20]

RecBole: Towards a Unified, Comprehensive and Efficient Framework for Recommendation Algorithms , author=
[21]

Proceedings of the 17th International Conference on Advances in Mobile Computing & Multimedia , pages=

Context-Aware Smart Energy Recommender (CASER) , author=. Proceedings of the 17th International Conference on Advances in Mobile Computing & Multimedia , pages=
[22]

Global Context Enhanced Graph Neural Networks for Session-based Recommendation , booktitle =

Ziyang Wang and Wei Wei and Gao Cong and Xiao. Global Context Enhanced Graph Neural Networks for Session-based Recommendation , booktitle =
[23]

Ruihong Qiu and Jingjing Li and Zi Huang and Hongzhi Yin , title =
[24]

Feng Yu and Yanqiao Zhu and Qiang Liu and Shu Wu and Liang Wang and Tieniu Tan , title =
[25]

Handling Information Loss of Graph Neural Networks for Session-based Recommendation , booktitle =

Tianwen Chen and Raymond Chi. Handling Information Loss of Graph Neural Networks for Session-based Recommendation , booktitle =
[26]

Proceedings of the 2017 ACM on Conference on Information and Knowledge Management , pages=

Neural attentive session-based recommendation , author=. Proceedings of the 2017 ACM on Conference on Information and Knowledge Management , pages=

2017
[27]

Proceedings of the 28th ACM international conference on information and knowledge management , pages=

Rethinking the item order in session-based recommendation with graph neural networks , author=. Proceedings of the 28th ACM international conference on information and knowledge management , pages=
[28]

Zhiqiang Pan and Fei Cai and Wanyu Chen and Honghui Chen and Maarten de Rijke , title =
[29]

S3-Rec: Self-Supervised Learning for Sequential Recommendation with Mutual Information Maximization , booktitle =

Kun Zhou and Hui Wang and Wayne Xin Zhao and Yutao Zhu and Sirui Wang and Fuzheng Zhang and Zhongyuan Wang and Ji. S3-Rec: Self-Supervised Learning for Sequential Recommendation with Mutual Information Maximization , booktitle =
[30]

Xin Xia and Hongzhi Yin and Junliang Yu and Qinyong Wang and Lizhen Cui and Xiangliang Zhang , title =
[31]

Junsu Cho and SeongKu Kang and Dongmin Hyun and Hwanjo Yu , title =
[32]

arXiv preprint arXiv:2010.14395 , year=

Contrastive Learning for Sequential Recommendation , author=. arXiv preprint arXiv:2010.14395 , year=

2010 arXiv
[33]

Huachi Zhou and Qiaoyu Tan and Xiao Huang and Kaixiong Zhou and Xiaoling Wang , title =
[34]

Bartlomiej Twardowski and Pawel Zawistowski and Szymon Zaborowski , title =
[35]

Revisiting Alternative Experimental Settings for Evaluating Top-N Item Recommendation Algorithms , booktitle =

Wayne Xin Zhao and Junhua Chen and Pengfei Wang and Qi Gu and Ji. Revisiting Alternative Experimental Settings for Evaluating Top-N Item Recommendation Algorithms , booktitle =
[36]

Filter-enhanced

Kun Zhou and Hui Yu and Wayne Xin Zhao and Ji. Filter-enhanced
[37]

2021 , booktitle =

Locally constrained self-attentive sequential recommendation , author =. 2021 , booktitle =

2021
[38]

Advances in neural information processing systems , volume=

Mlp-mixer: An all-mlp architecture for vision , author=. Advances in neural information processing systems , volume=
[39]

arXiv preprint arXiv:2104.05707 , year=

Localvit: Bringing locality to vision transformers , author=. arXiv preprint arXiv:2104.05707 , year=

arXiv
[40]

arXiv preprint arXiv:2107.08391 , year=

As-mlp: An axial shifted mlp architecture for vision , author=. arXiv preprint arXiv:2107.08391 , year=

arXiv
[41]

Advances in neural information processing systems , volume=

Global filter networks for image classification , author=. Advances in neural information processing systems , volume=
[42]

Advances in Neural Information Processing Systems , volume=

The Boltzmann perceptron network: a multi-layered feed-forward network equivalent to the Boltzmann machine , author=. Advances in Neural Information Processing Systems , volume=
[43]

Englewood Cliffs , year=

Continuous and discrete signals and systems , author=. Englewood Cliffs , year=
[44]

Englewood Cliffs: Prentice-Hall , year=

Theory and application of digital signal processing , author=. Englewood Cliffs: Prentice-Hall , year=
[45]

Contrastive Curriculum Learning for Sequential User Behavior Modeling via Data Augmentation , author=
[46]

Zihan Lin and Changxin Tian and Yupeng Hou and Wayne Xin Zhao , title=
[47]

Proceedings of the 45th international ACM SIGIR conference on research and development in information retrieval , pages=

Core: simple and effective session-based recommendation within consistent representation space , author=. Proceedings of the 45th international ACM SIGIR conference on research and development in information retrieval , pages=
[48]

Multiplex memory network for collaborative filtering , author=
[49]

Weinberger and John Blitzer and Lawrence K

Kilian Q. Weinberger and John Blitzer and Lawrence K. Saul , title =
[50]

, author=

Visualizing data using t-SNE. , author=
[51]

Brian Kulis , title =. Found. Trends Mach. Learn. , volume =
[52]

Hinton and Alex Krizhevsky and Ilya Sutskever and Ruslan Salakhutdinov , title =

Nitish Srivastava and Geoffrey E. Hinton and Alex Krizhevsky and Ilya Sutskever and Ruslan Salakhutdinov , title =
[53]

Kingma and Jimmy Ba , title =

Diederik P. Kingma and Jimmy Ba , title =
[54]

Kihyuk Sohn , title =
[55]

Gomez and Lukasz Kaiser and Illia Polosukhin , title =

Ashish Vaswani and Noam Shazeer and Niki Parmar and Jakob Uszkoreit and Llion Jones and Aidan N. Gomez and Lukasz Kaiser and Illia Polosukhin , title =
[56]

Belongie and Ser

Kevin Musgrave and Serge J. Belongie and Ser. A Metric Learning Reality Check , booktitle =
[57]

arXiv preprint arXiv:2104.08821 , year =

Tianyu Gao and Xingcheng Yao and Danqi Chen , title =. arXiv preprint arXiv:2104.08821 , year =

arXiv
[58]

Raia Hadsell and Sumit Chopra and Yann LeCun , title =
[59]

Girshick , title =

Kaiming He and Haoqi Fan and Yuxin Wu and Saining Xie and Ross B. Girshick , title =
[60]

ACM Computing Surveys (CSUR) , volume=

A survey on session-based recommender systems , author=. ACM Computing Surveys (CSUR) , volume=. 2021 , publisher=

2021
[61]

Proceedings of the IEEE conference on computer vision and pattern recognition , pages=

Deep residual learning for image recognition , author=. Proceedings of the IEEE conference on computer vision and pattern recognition , pages=
[62]

arXiv preprint arXiv:1607.06450 , year=

Layer normalization , author=. arXiv preprint arXiv:1607.06450 , year=

arXiv
[63]

Hinton , title =

Ting Chen and Simon Kornblith and Mohammad Norouzi and Geoffrey E. Hinton , title =
[64]

Tongzhou Wang and Phillip Isola , title =
[65]

Yuning You and Tianlong Chen and Yongduo Sui and Ting Chen and Zhangyang Wang and Yang Shen , title =
[66]

Proceedings of the 13th international conference on web search and data mining , pages=

Time interval aware self-attention for sequential recommendation , author=. Proceedings of the 13th international conference on web search and data mining , pages=
[67]

Proceedings of the 25th ACM SIGKDD international conference on knowledge discovery & data mining , pages=

Hierarchical gating networks for sequential recommendation , author=. Proceedings of the 25th ACM SIGKDD international conference on knowledge discovery & data mining , pages=
[68]

Computer , volume=

Matrix factorization techniques for recommender systems , author=. Computer , volume=. 2009 , publisher=

2009
[69]

Proceedings of the 10th international conference on World Wide Web , pages=

Item-based collaborative filtering recommendation algorithms , author=. Proceedings of the 10th international conference on World Wide Web , pages=
[70]

, author=

An MDP-based recommender system. , author=. Journal of Machine Learning Research , volume=
[71]

ACM computing surveys (CSUR) , volume=

Deep learning based recommender system: A survey and new perspectives , author=. ACM computing surveys (CSUR) , volume=. 2019 , publisher=

2019
[72]

arXiv preprint arXiv:1511.06939 , year=

Session-based recommendations with recurrent neural networks , author=. arXiv preprint arXiv:1511.06939 , year=

arXiv
[73]

Proceedings of the 1st workshop on deep learning for recommender systems , pages=

Improved recurrent neural networks for session-based recommendations , author=. Proceedings of the 1st workshop on deep learning for recommender systems , pages=
[74]

Proceedings of the eleventh ACM international conference on web search and data mining , pages=

Personalized top-n sequential recommendation via convolutional sequence embedding , author=. Proceedings of the eleventh ACM international conference on web search and data mining , pages=
[75]

Expert Systems with Applications , volume=

Session-based recommendation with time-aware neural attention network , author=. Expert Systems with Applications , volume=
[76]

, author=

Visualizing data using t-SNE. , author=. Journal of machine learning research , volume=
[77]

Proceedings of the 14th ACM International Conference on Web Search and Data Mining , pages=

Sparse-interest network for sequential recommendation , author=. Proceedings of the 14th ACM International Conference on Web Search and Data Mining , pages=
[78]

, author=

Graph Contextualized Self-Attention Network for Session-based Recommendation. , author=. IJCAI , volume=
[79]

The 41st international ACM SIGIR conference on research & development in information retrieval , pages=

Improving sequential recommendation with knowledge-enhanced memory networks , author=. The 41st international ACM SIGIR conference on research & development in information retrieval , pages=
[80]

arXiv preprint arXiv:1907.00590 , year=

A review-driven neural model for sequential recommendation , author=. arXiv preprint arXiv:1907.00590 , year=

1907 arXiv
[81]

Proceedings of the 44th international ACM SIGIR conference on research and development in information retrieval , pages=

Counterfactual data-augmented sequential recommendation , author=. Proceedings of the 44th international ACM SIGIR conference on research and development in information retrieval , pages=
[82]

Proceedings of the 44th International ACM SIGIR Conference on Research and Development in Information Retrieval , pages=

Causerec: Counterfactual user sequence synthesis for sequential recommendation , author=. Proceedings of the 44th International ACM SIGIR Conference on Research and Development in Information Retrieval , pages=
[83]

Proceedings of the Web Conference 2021 , pages=

A novel macro-micro fusion network for user representation learning on mobile apps , author=. Proceedings of the Web Conference 2021 , pages=

2021
[84]

Proceedings of the 29th ACM international conference on information & knowledge management , pages=

S3-rec: Self-supervised learning for sequential recommendation with mutual information maximization , author=. Proceedings of the 29th ACM international conference on information & knowledge management , pages=

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.

[1] [1]

Konstan and John Riedl , title =

Badrul Munir Sarwar and George Karypis and Joseph A. Konstan and John Riedl , title =

[2] [2]

Steffen Rendle and Christoph Freudenthaler and Zeno Gantner and Lars Schmidt

[3] [3]

Factorizing personalized Markov chains for next-basket recommendation , booktitle =

Steffen Rendle and Christoph Freudenthaler and Lars Schmidt. Factorizing personalized Markov chains for next-basket recommendation , booktitle =

[4] [4]

Session-based Recommendations with Recurrent Neural Networks , booktitle =

Bal. Session-based Recommendations with Recurrent Neural Networks , booktitle =

[5] [5]

Parallel Recurrent Neural Network Architectures for Feature-rich Session-based Recommendations , booktitle =

Bal. Parallel Recurrent Neural Network Architectures for Feature-rich Session-based Recommendations , booktitle =

work page

[6] [6]

Neural Collaborative Filtering , booktitle =

Xiangnan He and Lizi Liao and Hanwang Zhang and Liqiang Nie and Xia Hu and Tat. Neural Collaborative Filtering , booktitle =

work page

[7] [7]

Jing Li and Pengjie Ren and Zhumin Chen and Zhaochun Ren and Tao Lian and Jun Ma , title =

work page

[8] [8]

Qiao Liu and Yifu Zeng and Refuoe Mokhosi and Haibin Zhang , title =

work page