Pith. sign in

REVIEW 4 major objections 6 minor 1 cited by

Sparser Training for On-Device Recommendation Systems

T0 review · 4 major / 6 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read The paper claims that dynamic sparse training for recommendation embeddings can be made fully sparse in both forward and backward passes, and that the resulting method reaches state-of-the-art Recall@20 and NDCG@20 in 16 of 18 settings on…

desk verdict A credible empirical DST method for on-device collaborative filtering with a few evaluation gaps, most notably a missing uniform-sampling control and an unspecified omega in the main results. read the letter →

arxiv 2411.12205 v1 pith:XD24P4EC submitted 2024-11-19 cs.IR

Yunke Qu , Liang Qu , Tong Chen , Xiangyu Zhao , Jianxin Li , Hongzhi Yin This is my paper · ORCID
classification cs.IR
keywords RecommenderSystemsModelPruningCollaborativeFilteringDynamicSparseTrainingEmbeddingTablesNonnegativeMatrixFactorizationGradientsOn-DeviceRecommendation
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

Recommender systems on phones and other memory-limited devices are dominated by the embedding table, which stores a dense vector for every user and item. The paper argues that dynamic sparse training, which prunes most of that table, has three correctable weaknesses: the initial mask is random, regrowth decisions come from a single batch, and backpropagation still computes dense gradients. SparseRec replaces each piece: a nonnegative matrix factorization of the interaction matrix supplies the initial mask, cumulative gradients drive regrowth, and frequency-based sampling decides which vectors get gradients, so sparsity holds in both forward and backward passes. If these claims hold, a recommender can be trained on-device at a fraction of the memory cost while matching or beating heavier baselines. On two public datasets and three base models, the paper reports the best metric in 16 of 18 settings and up to 11.79% relative gain over the second-best baseline.

What carries the argument

The argument rides on three pieces of machinery. The first is a binary mask built by binarizing the nonnegative matrix factorization (NMF) of the interaction matrix, which supplies a naturally sparse, data-driven initialization instead of a random uniform mask. The second is a cumulative gradient table that sums sparse gradients over the exploration interval and selects regrown parameters by top-k magnitude, so reactivation reflects behavior over many batches. The third is multinomial sampling over user and item frequencies, which chooses which embedding rows receive gradient computation at all, leaving the backward pass sparse. A short memory analysis ties these together: with density $d$ and sampling ratio $\omega$, peak training storage is at most $(2d+2\omega)s(m+n)$ parameters, which is below the $(d+1)s(m+n)$ cost of conventional DST exactly when $\omega \le (1-d)/2$. That inequality is what guarantees the method uses less memory than dense-gradient DST.

What would settle it

Run SparseRec on a long-tail-heavy dataset with a dense-gradient oracle that keeps the NMF mask and cumulative regrowth but computes all gradients, and compare Recall@20 at the same density; a large gap would show the sparse-gradient sampling is the bottleneck, while swapping the frequency-based sampler for uniform row sampling would isolate whether frequency weighting or plain stochastic exploration drives the gains.

Watch

Extended reading notes

Core claim

The central claim is that a dynamic sparse training pipeline for graph-based recommenders can be made genuinely sparse in both directions of training without sacrificing accuracy, provided three mechanisms are changed together. SparseRec initializes the binary pruning mask by binarizing a nonnegative matrix factorization of the user-item interaction matrix, giving the sparse structure a data-driven starting point rather than a random one. It regrows inactive parameters according to gradients accumulated over the exploration window, so reactivated parameters serve users and items that matter across many batches rather than only the current one. It computes gradients only for active parameters plus a frequency-sampled subset of embedding rows, which keeps the backward pass sparse. The paper reports that this combination reduces peak memory relative to dense-gradient DST whenever the sampling ratio satisfies $\omega \le (1-d)/2$, and delivers the best metric in 16 of the 18 evaluated settings.

Load-bearing premise

The load-bearing premise is that a frequency-weighted random sample of embedding rows reveals which inactive parameters are most valuable to reactivate; if the important dormant parameters live in rows belonging to rare users or items, the sampling can miss them and regrowth slides toward random selection.

Editorial extensions

If this is right

  • Sparse training can run end to end on-device: the backward pass no longer allocates a dense gradient table, because gradients are computed only for active parameters plus the frequency-sampled rows.
  • Peak training memory is bounded by $(2d+2\omega)s(m+n)$ parameters, which is less than the $(d+1)s(m+n)$ cost of conventional DST whenever the sampling ratio satisfies $\omega \le (1-d)/2$.
  • NMF-based mask initialization outperforms random uniform initialization in the paper's ablations under both cumulative and instantaneous regrowth, so the starting structure of the sparse network matters.
  • Cumulative-gradient regrowth assigns larger effective embeddings to frequent users and items, with Pearson correlations around 0.8, while instantaneous gradients produce a weaker frequency-size alignment.
  • The method's largest relative improvement over the best baseline appears at the lowest density ratio (6.25%), suggesting the sparsest regimes benefit most from targeted regrowth.

Reading between the lines

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

  • A uniform-sampling ablation would separate the contribution of frequency weighting from plain stochastic exploration; the paper's ablations vary initialization and gradient accumulation but not the sampling distribution.
  • Because cumulative gradients push capacity toward frequent users and items, SparseRec could be combined with variable-size embedding methods that let the model decide where to spend dimensions, rather than fixing all rows to the same width.
  • The same recipe of data-driven mask, accumulated-gradient regrowth, and row-sampled sparse backward passes may generalize to other embedding-heavy models such as sequential or categorical-feature recommenders, although the paper only evaluates graph-based collaborative filtering.
  • In applications where rare users or items carry the most information, the frequency-based sampler is a liability; a salience-based sampler would be a natural variant the paper does not explore.
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.

Referee Report

4 major / 6 minor

Summary. The paper proposes SparseRec, a dynamic sparse training (DST) method for embedding tables in GNN-based recommender systems targeting on-device deployment. It makes three contributions: (1) initializing the binary mask matrix via Nonnegative Matrix Factorization (NMF) on the user-item interaction matrix; (2) regrowing pruned parameters according to cumulative gradients accumulated over several training steps rather than instantaneous gradients; and (3) computing gradients only for active parameters plus a frequency-weighted random sample of embedding vectors, thereby keeping the backward pass sparse. The method is evaluated on Gowalla and Yelp with XSimGCL, MGDCF, and LightGCN as base recommenders at three density ratios, reporting top scores in 16 of 18 settings and up to 11.79% relative improvement over the second-best baseline, with a theoretical memory-complexity advantage over prior DST methods.

Significance. SparseRec addresses a practically important problem—memory-efficient, on-device recommendation—and the experimental evaluation is reasonably broad: two datasets, three base recommenders, three density ratios, and several established baselines. If the method is sound as described, the NMF initialization and cumulative-gradient regrowth would be useful additions to the DST literature for recommender systems. The paper also includes a careful memory-complexity analysis (Section 4.4) that separates it from earlier dense-gradient DST methods. However, the central sparse-gradient mechanism needs a crucial clarification about how gradients for masked-out (inactive) parameters are obtained, and several experimental details (the sampling ratio used in the main results, run-level statistics) are missing. These issues do not render the empirical claims impossible, but they must be addressed before the method can be assessed as reproducible and fully supported.

major comments (4)
  1. [Section 4.2, Eq. (9)–(11); Algorithm 1] This is a load-bearing correctness concern because the sparse-gradient and regrowth mechanisms are the core contributions. If the gradients for inactive parameters are indeed zero, the reported performance improvements must be attributed to some other, unspecified mechanism.
  2. [Section 5.5.2 and Table 1] The memory advantage is a central selling point of the paper, so the missing ω makes the main memory claim unverifiable.
  3. [Section 5.3] Given that the paper claims superiority in 16/18 settings, this statistical evidence is load-bearing for the empirical contribution.
  4. [Table 2 and Section 4.2, Eqs. (7)–(8)] This ablation is important because the sparse-gradient mechanism is a claimed contribution; without it, the reader cannot tell whether the frequency-based sampling is justified or merely an arbitrary choice that happens not to hurt performance.
minor comments (6)
  1. [Abstract and Introduction] Typos: 'they typically initializes the mask matrix' should be 'they typically initialize the mask matrix'; 'an lightweight embedding method' should be 'a lightweight embedding method'.
  2. [Section 5.5] The heading 'Hyerparameter' appears twice and should be 'Hyperparameter'.
  3. [Eq. (7)] The notation 'p^U / p^U' appears to be a typo; it should likely be 'p^U / p^V' for the item sampling distribution.
  4. [Figure 3] The y-axis label 'User emb. size' appears on the item-frequency panels; it should read 'Item emb. size'.
  5. [Section 5.5.2] The claim that 'an ω close to zero essentially results in random regrowth' is imprecise: with tiny ω, only a very small candidate set is available, but the sampling is still frequency-weighted, not uniform random.
  6. [Eq. (12)] The regrowth selection should explicitly exclude active parameters from the top-k over |C|, since G is required to be disjoint from A (as shown in Figure 2). The text should state that C is zeroed or filtered for active parameters before applying top-k.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity: SparseRec's gains rest on controlled ablations and holdout evaluation, and the frequency-size correlation is measured with the sampling confound removed (omega=1), so no prediction reduces to its inputs by construction.

full rationale

SparseRec is an empirical method paper. The three claimed contributions (NMF-based mask initialization, cumulative-gradient regrowth, and sparse gradient sampling) are algorithm design choices whose value is established by controlled comparisons in Table 2 and by holdout metrics in Table 1, not by a chain of definitions that assumes its own conclusion. The frequency-weighted sampling in Eqs. 7-8 does determine which rows can accumulate gradients in Eq. 11 and therefore which inactive parameters are eligible for regrowth in Eq. 12, but that is the algorithm's intended mechanism, not a hidden circular step: the paper never claims the frequency-weighting is derived from the frequency-size correlation. Moreover, Section 5.4.2 explicitly states that to measure the frequency-size correlation 'we set the sampling ratio to 1 and use a dense cumulative gradient table,' so the Figure 3 correlations are not manufactured by Eq. 7-8. The memory comparison in Eqs. 13-15 is conditional algebra reducing to the bound omega <= (1-d)/2; it is a stated inequality, not a fitted parameter relabeled as a prediction. The NMF initialization does use the training interaction matrix R to produce the mask, but the mask is a training-time structural choice and all reported ranking metrics are computed on held-out test interactions, so this is legitimate use of training data rather than definitional circularity. The ablations in Table 2 vary mask initialization (NMF vs uniform) and gradient accumulation (cumulative vs instantaneous) while holding the rest of the pipeline fixed, and the main SOTA claim is evaluated across 18 settings on two datasets. The paper does contain several self-citations (e.g., [31-33, 39, 53]), but these appear as related-work context and are not load-bearing justifications for the paper's central claims. Additional baselines, such as uniform random sampling of embedding rows or a dense-gradient oracle, would strengthen the evaluation, but their absence is a completeness concern, not evidence that any result is identical to its input by construction. Overall, no specific prediction or derivation in the paper reduces to its own inputs, so the circularity burden is minimal.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

The method rests on a small set of domain assumptions inherited from the DST and embedding-pruning literature, plus three free hyperparameters. No new physical or mathematical entities are introduced.

free parameters (3)
  • Sampling ratio omega = not explicitly stated for main experiments; explored in {h, h/2, h/4, h/8, h/16} where h=(1-d)/2
    Controls how many embedding vectors are sampled for gradient computation and regrowth; affects memory and performance (Section 5.5.2).
  • Weight exploration interval Delta_T = 5b on Gowalla, 7b on Yelp
    Controls frequency of prune-regrow cycles; chosen based on Figure 4 (Section 5.5.1).
  • Initial pruning rate rho_0 = not specified
    Used in Eq. 4 for cosine annealing of the pruning rate; no value is reported in the paper.
assumptions (4)
  • domain assumption DSL-style DST with consistent sparsity can match dense model performance
    The paper builds on DSL[44] and the DST literature; this assumption underlies the entire approach (Section 3).
  • domain assumption NMF factorization of the interaction matrix yields a mask support aligned with important embedding dimensions
    Section 4.1 uses NMF on R to derive the mask M; the effectiveness of this initialization is assumed, not proven.
  • domain assumption User/item frequency is a valid proxy for the importance of their embedding parameters
    Section 4.2 samples vectors by frequency to guide sparse gradients and regrowth; supported by prior work [22,23] but assumed here.
  • domain assumption The base recommenders' hyperparameters inherited from original papers are optimal for the sparse setting
    Section 5.2: 'we inherit their optimal settings as reported in their original papers.'

how reviews work

0 comments
Cite this review

Pith. "Pith review of Sparser Training for On-Device Recommendation Systems." pith.science (2026). https://pith.science/paper/XD24P4EC

@misc{pith2026241112205,
  author       = {Pith},
  title        = {Pith review of: Sparser Training for On-Device Recommendation Systems},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/XD24P4EC}},
  note         = {Machine review of arXiv:2411.12205}
}
read the original abstract

Recommender systems often rely on large embedding tables that map users and items to dense vectors of uniform size, leading to substantial memory consumption and inefficiencies. This is particularly problematic in memory-constrained environments like mobile and Web of Things (WoT) applications, where scalability and real-time performance are critical. Various research efforts have sought to address these issues. Although embedding pruning methods utilizing Dynamic Sparse Training (DST) stand out due to their low training and inference costs, consistent sparsity, and end-to-end differentiability, they face key challenges. Firstly, they typically initializes the mask matrix, which is used to prune redundant parameters, with random uniform sparse initialization. This strategy often results in suboptimal performance as it creates unstructured and inefficient connections. Secondly, they tend to favor the users/items sampled in the single batch immediately before weight exploration when they reactivate pruned parameters with large gradient magnitudes, which does not necessarily improve the overall performance. Thirdly, while they use sparse weights during forward passes, they still need to compute dense gradients during backward passes. In this paper, we propose SparseRec, an lightweight embedding method based on DST, to address these issues. Specifically, SparseRec initializes the mask matrix using Nonnegative Matrix Factorization. It accumulates gradients to identify the inactive parameters that can better improve the model performance after activation. Furthermore, it avoids dense gradients during backpropagation by sampling a subset of important vectors. Gradients are calculated only for parameters in this subset, thus maintaining sparsity during training in both forward and backward passes.

Figures

Figures reproduced from arXiv: 2411.12205 by the authors.

Figure 1
Figure 1. A toy example of sparseRec. All parameters in E A (a) Before exploration All parameters in E A P (b) Pruning All parameters in E A P S (c) Sampling All parameters in E A P S G (d) Regrowing All parameters in E A (e) After exploration [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Venn diagrams illustrating the relationship between [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. The relationship between frequency and embed [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figures from the paper (1 more)
Figure 6
Figure 6. Figure 6: Hyerparameter analysis of the sampling ratio [PITH_FULL_IMAGE:figures/full_fig_p008_6.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

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

  1. On-device Content-based Recommendation with Single-shot Embedding Pruning: A Cooperative Game Perspective

    cs.IR 2024-11 conditional novelty 6.0 of 10

    Shaver prunes CTR model embedding tables to arbitrary sparsity in a single shot using Shapley-value attribution and a field-aware codebook, with accuracy close to retrained baselines.

Reference graph

Works this paper leans on

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

  1. [1]

    Hongxu Chen, Hongzhi Yin, Tong Chen, Weiqing Wang, Xue Li, and Xia Hu. 2022. Social Boosted Recommendation With Folded Bipartite Network Embedding. IEEE Transactions on Knowledge and Data Engineering 34, 2 (2022), 914–926

    work page 2022

  2. [2]

    Tim Dettmers and Luke Zettlemoyer. 2019. Sparse Networks from Scratch: Faster Training without Losing Performance. arXiv:1907.04840

    arXiv 2019

  3. [3]

    Utku Evci, Trevor Gale, Jacob Menick, Pablo Samuel Castro, and Erich Elsen

  4. [4]

    Jonathan Frankle and Michael Carbin. 2019. The Lottery Ticket Hypothesis: Finding Sparse, Trainable Neural Networks. In ICLR

    2019

  5. [5]

    Song Han, Jeff Pool, John Tran, and William J. Dally. 2015. Learning both weights and connections for efficient neural networks. In NeurIPS. 1135–1143

    work page 2015

  6. [6]

    Xiangnan He, Kuan Deng, Xiang Wang, Yan Li, Yongdong Zhang, and Meng Wang. 2020. Lightgcn: Simplifying and powering graph convolution network for recommendation. In SIGIR. 639–648

    2020

  7. [7]

    Xiangnan He, Lizi Liao, Hanwang Zhang, Liqiang Nie, Xia Hu, and Tat-Seng Chua. 2017. Neural Collaborative Filtering. In WWW. 173–182

    2017

  8. [8]

    Xiangnan He, Hanwang Zhang, Min-Yen Kan, and Tat-Seng Chua. 2016. Fast Matrix Factorization for Online Recommendation with Implicit Feedback. In SIGIR. 549–558

    work page 2016

Show all 67 references
  1. [9]

    Mike Heddes, Narayan Srinivasa, Tony Givargis, and Alexandru Nicolau. 2024. Always-Sparse Training by Growing Connections with Guided Stochastic Explo- ration. arXiv:2401.06898

    2024 arXiv

  2. [10]

    Jun Hu, Bryan Hooi, Shengsheng Qian, Quan Fang, and Changsheng Xu. 2022. MGDCF: Distance Learning via Markov Graph Diffusion for Neural Collaborative Filtering. IEEE Transactions on Knowledge and Data Engineering 36 (2022), 3281– 3296

    2022

  3. [11]

    Yifan Hu, Yehuda Koren, and Chris Volinsky. 2008. Collaborative Filtering for Implicit Feedback Datasets. In ICDM. 263–272

    2008

  4. [12]

    Janowsky

    Steven A. Janowsky. 1989. Pruning versus clipping in neural networks. Phys. Rev. A 39 (1989), 6600–6603

    1989

  5. [13]

    Wang-Cheng Kang, Derek Zhiyuan Cheng, Tiansheng Yao, Xinyang Yi, Ting Chen, Lichan Hong, and Ed H. Chi. 2021. Learning to Embed Categorical Features without Embedding Tables for Recommendation. In KDD. 840–850

    2021

  6. [14]

    Diederik P Kingma. 2014. Adam: A method for stochastic optimization. arXiv preprint arXiv:1412.6980 (2014)

    2014 arXiv

  7. [15]

    Yehuda Koren, Steffen Rendle, and Robert Bell. 2021. Advances in collaborative filtering. Recommender systems handbook (2021), 91–142

    2021

  8. [16]

    Yann LeCun, John Denker, and Sara Solla. 1989. Optimal Brain Damage. In NeurIPS, D. Touretzky (Ed.), Vol. 2. 598–605

    1989

  9. [17]

    Joonseok Lee, Sami Abu-El-Haija, Balakrishnan Varadarajan, and Apostol Natsev

  10. [18]

    Namhoon Lee, Thalaiyasingam Ajanthan, and Philip Torr. 2019. SNIP: SINGLE- SHOT NETWORK PRUNING BASED ON CONNECTION SENSITIVITY. InICLR

    2019

  11. [19]

    Jiacheng Li, Ming Wang, Jin Li, Jinmiao Fu, Xin Shen, Jingbo Shang, and Julian McAuley. 2023. Text Is All You Need: Learning Language Representations for Sequential Recommendation. In KDD. 1258–1267

    2023

  12. [20]

    Xurong Liang, Tong Chen, Lizhen Cui, Yang Wang, Meng Wang, and Hongzhi Yin. 2024. Lightweight Embeddings for Graph Collaborative Filtering. In SIGIR

    2024

  13. [21]

    Xurong Liang, Tong Chen, Quoc Viet Hung Nguyen, Jianxin Li, and Hongzhi Yin

  14. [22]

    Haochen Liu, Xiangyu Zhao, Chong Wang, Xiaobing Liu, and Jiliang Tang. 2020. Automated Embedding Size Search in Deep Recommender Systems. In SIGIR. 2307–2316

    2020

  15. [23]

    Siyi Liu, Chen Gao, Yihong Chen, Depeng Jin, and Yong Li. 2021. Learnable Embedding sizes for Recommender Systems. In ICLR

    2021

  16. [24]

    Mocanu, and Mykola Pechenizkiy

    Shiwei Liu, Lu Yin, Decebal C. Mocanu, and Mykola Pechenizkiy. 2021. Do We Actually Need Dense Over-Parameterization? In-Time Over-Parameterization in Sparse Training. In ICML. 6989–7000

    2021

  17. [25]

    Kelong Mao, Jieming Zhu, Xi Xiao, Biao Lu, Zhaowei Wang, and Xiuqiang He

  18. [26]

    Decebal Constantin Mocanu, Elena Mocanu, Peter Stone, Phuong H Nguyen, Madeleine Gibescu, and Antonio Liotta. 2018. Scalable training of artificial neural networks with adaptive sparse connectivity inspired by network science. Nature communications 9, 1 (2018), 2383

    2018

  19. [27]

    Pavlo Molchanov, Arun Mallya, Stephen Tyree, Iuri Frosio, and Jan Kautz. 2019. Importance estimation for neural network pruning. In CVPR. 11264–11272

    2019

  20. [28]

    Pavlo Molchanov, Stephen Tyree, Tero Karras, Timo Aila, and Jan Kautz. 2017. Pruning Convolutional Neural Networks for Resource Efficient Inference. In International Conference on Learning Representations

    2017

  21. [29]

    Michael C Mozer and Paul Smolensky. 1989. Skeletonization: A Technique for Trimming the Fat from a Network via Relevance Assessment. InNeurIPS. 107–115

    1989

  22. [30]

    Hongwu Peng, Deniz Gurevin, Shaoyi Huang, Tong Geng, Weiwen Jiang, Orner Khan, and Caiwen Ding. 2022. Towards sparsification of graph neural networks. In 2022 IEEE 40th International Conference on Computer Design (ICCD) . 272–279

    2022

  23. [31]

    Yunke Qu, Tong Chen, Quoc Viet Hung Nguyen, and Hongzhi Yin. 2024. Budgeted Embedding Table For Recommender Systems. In WSDM. 557–566

    2024

  24. [32]

    Yunke Qu, Tong Chen, Xiangyu Zhao, Lizhen Cui, Kai Zheng, and Hongzhi Yin

  25. [33]

    Yunke Qu, Liang Qu, Tong Chen, Zhao Xiangyu, Nguyen Quoc Viet Hung, and Hongzhi Yin. 2024. Scalable Dynamic Embedding Size Search for Streaming Recommendation. In CIKM

    2024

  26. [34]

    Steffen Rendle, Christoph Freudenthaler, Zeno Gantner, and Lars Schmidt-Thieme

  27. [35]

    Hao-Jun Michael Shi, Dheevatsa Mudigere, Maxim Naumov, and Jiyan Yang

  28. [36]

    Nikko Ström. 1997. Sparse connection and pruning in large dynamic artificial neural networks. In Proc. 5th European Conference on Speech Communication and Technology (Eurospeech 1997). 2807–2810

    1997

  29. [37]

    In SIGIR

    Continuous Input Embedding Size Search For Recommender Systems. In SIGIR. 708–717

  30. [38]

    Georg Thimm and Emile Fiesler. 1995. Evaluating pruning methods. InProceedings of the International Symposium on Artificial neural networks . 20–25

    1995

  31. [39]

    Hung Vinh Tran, Tong Chen, Quoc Viet Hung Nguyen, Zi Huang, Lizhen Cui, and Hongzhi Yin. 2024. A Thorough Performance Benchmarking on Lightweight Embedding-based Recommender Systems. arXiv:2406.17335

    2024 arXiv

  32. [40]

    Chaoqi Wang, Guodong Zhang, and Roger Grosse. 2020. Picking Winning Tickets Before Training by Preserving Gradient Flow. In ICLR

    2020

  33. [41]

    Hao Wang, Yanmei Fu, Qinyong Wang, Hongzhi Yin, Changying Du, and Hui Xiong. 2017. A Location-Sentiment-Aware Recommender System for Both Home- Town and Out-of-Town Users. InKDD. 1135–1143

    2017

  34. [42]

    Compositional Embeddings Using Complementary Partitions for Memory- Efficient Recommendation Systems. In KDD. 165–175

  35. [43]

    Qinyong Wang, Hongzhi Yin, Tong Chen, Zi Huang, Hao Wang, Yanchang Zhao, and Nguyen Quoc Viet Hung. 2020. Next Point-of-Interest Recommendation on Resource-Constrained Mobile Devices. In WWW. 906–916

    2020

  36. [44]

    Hidenori Tanaka, Daniel Kunin, Daniel L. K. Yamins, and Surya Ganguli. 2020. Pruning neural networks without any data by iteratively conserving synaptic flow. In NeurIPS. 13 pages

    2020

  37. [45]

    Suhang Wang, Yilin Wang, Jiliang Tang, Kai Shu, Suhas Ranganath, and Huan Liu

  38. [46]

    Xiang Wang, Xiangnan He, Meng Wang, Fuli Feng, and Tat-Seng Chua. 2019. Neural graph collaborative filtering. In SIGIR. 165–174

    2019

  39. [47]

    Yuan Wang, Zhiyu Li, Changshuo Zhang, Sirui Chen, Xiao Zhang, Jun Xu, and Quan Lin. 2024. Do Not Wait: Learning Re-Ranking Model Without User Feedback At Serving Time in E-Commerce. In Proceedings of the 18th ACM Conference on Recommender Systems. 896–901

    2024

  40. [48]

    Yanfang Wang, Yongduo Sui, Xiang Wang, Zhenguang Liu, and Xiangnan He

  41. [50]

    Xin Xia, Junliang Yu, Qinyong Wang, Chaoqun Yang, Nguyen Quoc Viet Hung, and Hongzhi Yin. 2023. Efficient On-Device Session-Based Recommendation. ACM Trans. Inf. Syst. 41, 4 (2023), 24 pages

    2023

  42. [51]

    Shuyao Wang, Yongduo Sui, Jiancan Wu, Zhi Zheng, and Hui Xiong. 2024. Dy- namic Sparse Learning: A Novel Paradigm for Efficient Recommendation. In WSDM. 740–749

    2024

  43. [52]

    Hongzhi Yin, Bin Cui, Zi Huang, Weiqing Wang, Xian Wu, and Xiaofang Zhou

  44. [53]

    Hongzhi Yin, Liang Qu, Tong Chen, Wei Yuan, Ruiqi Zheng, Jing Long, Xin Xia, Yuhui Shi, and Chengqi Zhang. 2024. On-Device Recommender Systems: A Comprehensive Survey. arXiv:2401.11441

    2024 arXiv

  45. [54]

    Junliang Yu, Xin Xia, Tong Chen, Lizhen Cui, Nguyen Quoc Viet Hung, and Hongzhi Yin. 2024. XSimGCL: Towards Extremely Simple Graph Contrastive Learning for Recommendation. IEEE Transactions on Knowledge and Data Engi- neering 36, 2 (2024), 913–926

    2024

  46. [55]

    Changshuo Zhang, Sirui Chen, Xiao Zhang, Sunhao Dai, Weijie Yu, and Jun Xu. 2024. Reinforcing Long-Term Performance in Recommender Systems with User-Oriented Exploration Policy. In SIGIR. 1850–1860

    2024

  47. [56]

    Changshuo Zhang, Teng Shi, Xiao Zhang, Qi Liu, Ruobing Xie, Jun Xu, and Ji- Rong Wen. 2024. Modeling Domain and Feedback Transitions for Cross-Domain Conference acronym ’XX, June 03–05, 2018, Woodstock, NY Yunke Qu et al. Sequential Recommendation. arXiv:2408.08209

    2024 arXiv

  48. [57]

    Changshuo Zhang, Teng Shi, Xiao Zhang, Yanping Zheng, Ruobing Xie, Qi Liu, Jun Xu, and Ji-Rong Wen. 2024. QAGCF: Graph Collaborative Filtering for Q&A Recommendation. arXiv:2406.04828

    2024 arXiv

  49. [58]

    Xin Xia, Hongzhi Yin, Junliang Yu, Qinyong Wang, Guandong Xu, and Quoc Viet Hung Nguyen. 2022. On-Device Next-Item Recommendation with Self- Supervised Knowledge Distillation. In SIGIR. 546–555

    2022

  50. [60]

    Hongzhi Yin and Bin Cui. 2016. Spatio-Temporal Recommendation in Social Media (1st ed.). Springer Publishing Company, Incorporated

    2016

  51. [68]

    Ruiqi Zheng, Liang Qu, Tong Chen, Kai Zheng, Yuhui Shi, and Hongzhi Yin. 2024. Personalized Elastic Embedding Learning for On-Device Recommendation. IEEE Transactions on Knowledge and Data Engineering 36 (2024), 3363–3375

    2024

  52. [2009]

    In Proceedings of the Twenty-Fifth Conference on Uncertainty in Artificial Intelligence

    BPR: Bayesian personalized ranking from implicit feedback. In Proceedings of the Twenty-Fifth Conference on Uncertainty in Artificial Intelligence . 452–461

  53. [2015]

    In Proceedings of the 23rd ACM International Conference on Multimedia

    Joint Modeling of Users’ Interests and Mobility Patterns for Point-of-Interest Recommendation. In Proceedings of the 23rd ACM International Conference on Multimedia. 819–822

  54. [2017]

    What Your Images Reveal: Exploiting Visual Contents for Point-of-Interest Recommendation. In WWW. 391–400

  55. [2018]

    Collaborative deep metric learning for video understanding. In KDD. 481– 490

  56. [2020]

    In Proceedings of the 37th International Conference on Machine Learning

    Rigging the lottery: making all tickets winners. In Proceedings of the 37th International Conference on Machine Learning . 2943 – 2952

  57. [2021]

    UltraGCN: Ultra Simplification of Graph Convolutional Networks for Recommendation. In CIKM. 1253–1262

  58. [2022]

    Inter- national Journal of Intelligent Systems 37, 5 (2022), 3006–3024

    Exploring lottery ticket hypothesis in media recommender systems. Inter- national Journal of Intelligent Systems 37, 5 (2022), 3006–3024

    2022

  59. [2023]

    Learning compact compositional embeddings via regularized pruning for recommendation. In ICDM. 378–387

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.