Collaborative Diffusion Model for Recommender System

Gyuseok Lee; HwanJo Yu; Jundong Li; Yaochen Zhu; Yao Zhou

REVIEW 6 major objections 4 minor 30 references

Collaborative Diffusion Model for Recommender System

T0 review · 6 major / 4 minor · reviewed 2026-08-09 · deepseek-v4-flash

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

Pith's one-line read Neighbor signals folded into diffusion denoising lift top-N recommendation accuracy.

desk verdict A nice pseudo-user + neighbor blend for diffusion recommenders, but the math as written doesn't define how neighbor predictions are computed; the consistent gains are therefore hard to trust without a reproducibility pass. read the letter →

arxiv 2501.18997 v1 pith:55HRQ5MP submitted 2025-01-31 cs.IR

Gyuseok Lee , Yaochen Zhu , Hwanjo Yu , Yao Zhou , Jundong Li This is my paper · ORCID

classification cs.IR

keywords diffusionmodelrecommendersystemcollaborativefilteringitem-sideinformationpseudo-userneighborhood-basedrecommendationimplicitfeedbacktop-N

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

Diffusion-based recommender systems typically train a denoiser to reconstruct a user's binary interaction history from Gaussian noise, but heavy noise can wash out the personal detail that matters for recommendation. This paper claims that the lost personalized information can be recovered by blending the user's own denoised vector with the denoised vectors of two kinds of neighbors: real users with similar behavior, and 'pseudo-users' created from item review words. The paper introduces CDiff4Rec, which builds these neighbor sets by cosine similarity and incorporates their preference information directly into the diffusion reconstruction objective. On three public datasets (Yelp, Amazon-Game, Citeulike-t), the full model beats six baselines on Recall@20 and NDCG@20, with the largest gains over the plain diffusion recommender that it extends. If this holds, it means item content and collaborative signals can be folded into the diffusion denoising objective without changing the backbone architecture.

What carries the argument

The machinery is a modified diffusion reconstruction target that averages the denoiser's output for the query user with weighted outputs for behavior-similar real neighbors and review-word pseudo-neighbors. Pseudo-users are generated by treating each review word as a user whose interaction vector is the TF-IDF weighted, min-max normalized word co-occurrence across items, making the vector continuous in $[0,1]^{|I|}$. Top-$K$ neighbor sets for every user are precomputed with cosine distance and cached, so the extra cost during training and inference is just the aggregation itself. The same denoiser MLP that reconstructs the query user also produces the neighbor predictions, and the blended prediction replaces the plain reconstruction in the diffusion loss, which is what forces the model to encode collaborative and content signals.

What would settle it

Replace each neighbor's denoised prediction in Eq. (6) with that neighbor's observed (noiseless) interaction vector, keeping the same neighbor sets and blend weights; if Recall@20 does not drop to the level of the single-source ablations, then the improvement does not actually come from denoising the neighbors.

Watch

Extended reading notes

Core claim

CDiff4Rec's central claim is that the diffusion denoising objective should not stop at the user's own corrupted vector. Instead, the model defines a fine-grained reconstruction $\hat{\mathbf{r}}'_u$ as a weighted blend of the user's own denoiser output, the denoiser outputs of the top-$K$ most behaviorally similar real users, and the denoiser outputs of top-$K$ pseudo-users (review words with normalized TF-IDF interaction vectors). The weights $\alpha$, $\beta$, and $\gamma$ sum to one, and attention scores among neighbors can be average pooling, precomputed cosine similarity, or learned parametric attention. The training loss in Eq. (8) is the same diffusion loss as the base DR, but with $\hat{\mathbf{r}}'_u$ in place of the plain single-user reconstruction, so the denoiser is trained to make the blended prediction match the user's true interaction vector. The experiments show this single change improves Recall@20 and NDCG@20 over all baselines on all three datasets, and that both neighbor types contribute, with the largest gains when combined.

Load-bearing premise

The load-bearing premise is that a denoiser trained on binary user interaction vectors also produces accurate denoised predictions for real neighbors and for continuous TF-IDF pseudo-user vectors, since the final reconstruction blends those predictions and is trained to match the user's original vector.

Editorial extensions

If this is right

CDiff4Rec improves top-20 recall and NDCG over the strongest baselines on Yelp, AM-Game, and Citeulike-t, and both neighbor types contribute, with the largest gains when real and pseudo neighbors are combined.
The accuracy gain over DiffRec is consistent across R@10, R@50, and R@100 with modest wall-clock overhead (about 10% on Yelp and near-zero on AM-Game), so the collaborative blend is cheap to add.
Strong performance is reached with only 1,000 pseudo-users and top-20 neighbors, meaning the precomputed neighbor sets keep the method scalable.
Because pseudo-users come from item-side features, item content can be exploited without adding a new loss term or a separate network for content.
The choice among average, behavior-similarity, and parametric attention lets practitioners trade simplicity, efficiency, and effectiveness, with behavior-similarity attention giving the best reported results.

Reading between the lines

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

The pseudo-user recipe should generalize to any item-side feature source (tags, categories, visual attributes), since it only needs a feature-item co-occurrence matrix and a normalization step.
Because neighbor sets are precomputed, this collaborative blend could be grafted onto future denoising backbones without retraining the neighbor selection, making it a plug-in training objective.
A direct test of whether the benefit comes from the blended target or from the shared denoiser would be to compute neighbor predictions with a separate, frozen denoiser while training the query denoiser; the paper leaves this compartmentalization unspecified.
The reported sensitivity to $K$ and the number of pseudo-users suggests an automatic selection rule based on validation recall could squeeze out further gains, but that is not explored in the paper.

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

Quick read: the paper has a genuinely useful idea -- treating review words as pseudo-users and adding neighbor denoised predictions into the diffusion objective -- and the reported gains over DiffRec/Ease are consistent across three datasets. That part earns its keep. But the formal description is incomplete: Eq. (6) and Eq. (8) do not define a computable objective unless the authors specify how neighbor predictions are generated during training and inference.

What's new: this combination of pseudo-users from item features and neighbor aggregation inside the denoising target does not appear in the cited diffusion-recommender literature. The individual tricks are standard elsewhere (review words as users, kNN smoothing), but the integration is a legitimate extension of DiffRec. The ablations make sense: real neighbors alone add a bit, pseudo-users add a bit, and the combined model is best on all three datasets. Wall time is reported to increase only a little, which is credible if the neighbor sets are cached.

The soft spots are real. The biggest one is the unspecified protocol for \hat{r}_{r_u_i} and \hat{m}_{p_u_j} in Eq. (6). The denoiser takes a noisy vector and a timestep; each neighbor has its own input. Eq. (8) samples only the query user's x_t, with no joint noise distribution for the neighbors, no statement about whether neighbor predictions are computed on the fly or cached, and no description of an inference-time approximation. Running the diffusion chain for every neighbor at every step would be far more expensive than the reported 10% overhead, so something is missing. Second, pseudo-user vectors are continuous TF-IDF values in [0,1] while the denoiser is trained on binary user vectors; there is no explicit pseudo-user reconstruction loss. The model may learn to handle this implicitly, but that is not explained. Minor issues: the text says top-20 neighbors are best but Table 4 shows Yelp preferring top-50; alpha, beta, gamma values are not reported per dataset; and the paired t-tests lack procedural detail (number of runs, error bars). None of these are fatal by themselves, but together they prevent reproduction.

This is a five-page workshop-style paper with a nice idea and plausible, small effect sizes. It does not deserve rejection on the merits, but it does need a precise statement of the training/inference algorithm before the numbers can be accepted. Send it to serious review with a request for that specification and for the per-dataset hyperparameters.

Referee Report

6 major / 4 minor

Summary. The paper proposes CDiff4Rec, a diffusion-based collaborative filtering method that augments the query user's denoised interaction vector with the denoised vectors of two kinds of neighbors: top-K real users selected by cosine behavior similarity, and top-K pseudo-users constructed from review words by treating each word as a pseudo-user with TF-IDF-normalized interaction weights. The blended prediction is a convex combination with attention scores, and the training loss extends the DiffRec objective by replacing the denoised reconstruction with this blended vector. The paper reports Recall@20 and NDCG@20 gains over BPRMF, LightGCN, Ease, MultiVAE, ConVAE, and DiffRec on Yelp, Amazon-Game, and Citeulike-t, with ablations showing that both real and pseudo neighbors contribute, plus an accuracy-efficiency comparison against DiffRec.

Significance. If the method is fully specified and reproducible, the paper makes a useful contribution: it shows that item-content signals and collaborative neighbor signals can be folded directly into the diffusion objective without changing the backbone, and the pseudo-user construction is a simple and portable way to exploit review texts in a CF-oriented diffusion model. The consistent gains across three datasets, the component ablations, and the modest wall-clock overhead in Table 3 are genuine strengths. However, the central equations as written omit the sampling and inference protocol on which the headline gains depend, and the treatment of continuous pseudo-user vectors with a binary-trained denoiser is an unstated structural assumption, so the empirical claims cannot currently be checked from the text.

major comments (6)

[§3.3, Eq. (6), Eq. (8)] As written, the training objective is not computable: x̂_θ requires a noisy input and a timestep, so the neighbor terms ˆr_{ru_i} and ˆm_{pu_j} in Eq. (6) each require their own noisy vectors x_t^{(ru_i)} and x_t^{(pu_j)}, yet Eq. (8) samples only q(x_t | x0) for the query user and specifies no joint distribution over the neighbors' noisy inputs. Please state the sampling scheme used for neighbor inputs during training, or, if the query's x_t is reused for all neighbors, say so explicitly and justify this choice.
[§3.3, §4.1.3] The inference protocol for Eq. (6) is unspecified: computing each neighbor term requires running the denoiser on a noisy input for every neighbor, and the paper does not say whether this is done with a separate reverse chain per neighbor, with a shared noisy input, or with cached forward predictions. This matters for the wall-clock comparison in Table 3 and for the reproducibility of the gains in Table 2.
[§3.1, §3.3] The denoiser is trained on binary user vectors through Eq. (8), but pseudo-user vectors m_{pu} are continuous TF-IDF values normalized to [0,1]. The paper therefore relies on the unstated assumption that a denoiser trained on binary inputs generalizes to continuous pseudo-user inputs; please add a pseudo-user training term or provide evidence (e.g., a validation experiment) that this transfer works.
[§3.2, Eq. (5)] As written, the top-K real-neighbor set U_u^{ru} is selected from all users including the query user u itself. Because cosine distance between r_u and itself is zero, the query user would typically be its own nearest neighbor and would enter the blend in Eq. (6), potentially inflating the reported improvements. Please exclude the query user from the neighbor search or clarify if self-neighbors are intentionally retained.
[§2, Eq. (1)] The definition of \bar\alpha_t as a product over t'=1 to T makes \bar\alpha_t independent of t, which would break the forward process in Eq. (1) and the mean and loss formulas in Eqs. (2) and (4). This should be a product up to t; please correct the index and verify that the implementation uses the corrected definition.
[§4.1.3, Table 2] The actual values of α, β, γ, K, and |F| used for the reported main results are missing; the text gives only search ranges, and Table 4 shows non-monotonic sensitivity to K and |F|. In addition, the paired t-test is not described (unit of analysis, number of paired observations, and whether it is over users or repeated runs). Please report the selected hyperparameters and the test details so the significance claims can be checked.

minor comments (4)

[§3.2, Eq. (5)] The argsort notation does not state whether the order is ascending or descending; since distance is being sorted, please define the intended ordering explicitly.
[§4.3.2, Table 4] Table 4 does not state the reported metric in the caption; it should be identified (presumably R@20), and the caption of Table 3 should also spell out that the metrics are Recall and NDCG at the listed ranks.
[§4.3.2] The sentence claiming that top-20 neighbors and 1,000 pseudo-users are superior across the three datasets is not fully supported by Table 4, where Yelp achieves its best R@20 in the K=50 column for the 1,000-pseudo-user setting; please clarify the selection criterion or qualify the claim.
[§5] The conclusion repeats the abstract's phrase "loss of personalized information" as if it were a quantity; consider rephrasing to "mitigates the loss of personalized information" or "reduces the degradation of personalized information" for clarity.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the reported gains are measured against held-out interactions with standard validation-based hyperparameter tuning.

full rationale

The paper's central claim that blending the denoised query-user vector with denoised real and pseudo-neighbor vectors improves top-N recommendation (Eq. (6), Table 2) is an empirical result assessed on held-out test splits (Section 4.1.1), not a quantity equal to its own inputs by construction. Pseudo-users are TF-IDF item-feature aggregates (Section 3.1), neighbors are selected by cosine behavior similarity (Eq. (5)), and mixture weights alpha, beta, gamma plus neighborhood sizes are tuned on validation data (Section 4.1.3). This is standard supervised model selection, not a fitted parameter renamed as a prediction. The only same-group citation is [4] (Han et al., with co-author Hwanjo Yu), used merely as one example in a list of diffusion recommenders; it does not supply the architecture, a uniqueness theorem, or the experimental outcome, so it is not load-bearing. The underspecification of how neighbor noisy inputs are sampled in Eq. (6)/(8) is a reproducibility and correctness concern, but it does not make the improvement equal to the training objective or to the input features by construction. Hence no circular step is exhibited.

Assumptions & free parameters 5 free parameters · 3 assumptions · 1 invented entities

The central claim rests on the model architecture and training objective rather than on a mathematical derivation. The most important free parameters are the mixing weights alpha, beta, gamma and the neighbor count K; their fitted values are not reported. The pseudo-user construction is an ad hoc but explicit modeling choice. No new physical or latent entities are introduced beyond the deterministic pseudo-user representation.

free parameters (5)

alpha, beta, gamma = Not reported per dataset
Mixing weights for self, real neighbors, and pseudo neighbors in Eq. (6); searched over {0.1, 0.3, 0.5, 0.7, 0.9} with the constraint alpha+beta+gamma=1 (Section 4.1.3).
K (top-K neighbors) = Not reported per dataset
Number of real and pseudo neighbors in Eq. (5); searched over {10, 20, 50}. Table 4 shows different optimal values per dataset.
Number of pseudo-users |F| = Not reported per dataset
Number of review words used as pseudo-users, selected by highest average TF-IDF; studied at 1,000 to 20,000 in Table 4.
Diffusion timesteps T and noise schedule alpha_t = Inherited from DiffRec [20]
They follow the settings in DiffRec; the specific values are not given, so the central result depends on this external configuration.
TF-IDF and min-max normalization settings = Not specified
The construction of pseudo-user vectors in Section 3.1 depends on these preprocessing choices, but no parameters (e.g., sublinear tf, smooth idf) are reported.

assumptions (3)

ad hoc to paper A denoiser trained on binary user vectors generalizes to continuous TF-IDF pseudo-user vectors.
Real user vectors are binary (R in {0,1}) while pseudo-user vectors are continuous in [0,1]; Eq. (6) applies the same shared denoiser to both without domain adaptation (Sections 3.1 and 3.3).
domain assumption Cosine distance on observed binary interactions is a reliable and stable proxy for preference similarity.
Neighbors in Eq. (5) are chosen from the same binarized training interactions; the paper does not analyze sensitivity to the random 80/10/10 split or to data sparsity.
domain assumption The blended reconstruction target in Eq. (8) does not collapse to a trivial mean predictor of the training set.
Since r-hat'_u is a convex combination of user and neighbor predictions, the model could in principle minimize loss by averaging neighbors; the paper reports gains, implying this collapse does not happen, but provides no mechanism or analysis.

invented entities (1)

Pseudo-users independent evidence
purpose: Inject item-side content (review words) into the diffusion recommender as extra users with TF-IDF interaction vectors.
Pseudo-users are deterministic transformations of the item feature matrix and can be reconstructed by anyone with the data; they are not latent or unobservable.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Collaborative Diffusion Model for Recommender System." pith.science (2026). https://pith.science/paper/55HRQ5MP

@misc{pith2026250118997,
  author       = {Pith},
  title        = {Pith review of: Collaborative Diffusion Model for Recommender System},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/55HRQ5MP}},
  note         = {Machine review of arXiv:2501.18997}
}

read the original abstract

Diffusion-based recommender systems (DR) have gained increasing attention for their advanced generative and denoising capabilities. However, existing DR face two central limitations: (i) a trade-off between enhancing generative capacity via noise injection and retaining the loss of personalized information. (ii) the underutilization of rich item-side information. To address these challenges, we present a Collaborative Diffusion model for Recommender System (CDiff4Rec). Specifically, CDiff4Rec generates pseudo-users from item features and leverages collaborative signals from both real and pseudo personalized neighbors identified through behavioral similarity, thereby effectively reconstructing nuanced user preferences. Experimental results on three public datasets show that CDiff4Rec outperforms competitors by effectively mitigating the loss of personalized information through the integration of item content and collaborative signals.

Figures

Figures reproduced from arXiv: 2501.18997 by the authors.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

30 extracted references · 11 canonical work pages

[1]

Tommaso Carraro, Mirko Polato, Luca Bergamin, and Fabio Aiolli. 2022. Condi- tioned variational autoencoder for top-n item recommendation. In International Conference on Artificial Neural Networks . Springer, 785–796

work page 2022
[2]

Dong-Kyu Chae, Jin-Soo Kang, Sang-Wook Kim, and Jung-Tae Lee. 2018. CFGAN: A generic collaborative filtering framework based on generative adversarial networks. In Proceedings of the 27th ACM international conference on information and knowledge management. 137–146

work page 2018
[3]

Hanwen Du, Huanhuan Yuan, Zhen Huang, Pengpeng Zhao, and Xiaofang Zhou. 2023. Sequential recommendation with diffusion models. arXiv preprint arXiv:2304.04541 (2023)

arXiv 2023
[4]

Gwangseok Han, Wonbin Kweon, Minsoo Kim, and Hwanjo Yu. 2024. Controlling Diversity at Inference: Guiding Diffusion Recommender Models with Targeted Category Preferences. arXiv preprint arXiv:2411.11240 (2024)

work page Pith review arXiv 2024
[5]

GM Harshvardhan, Mahendra Kumar Gourisaria, Manjusha Pandey, and Sid- dharth Swarup Rautaray. 2020. A comprehensive survey and analysis of genera- tive models in machine learning. Computer Science Review 38 (2020), 100285

work page 2020
[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 Proceedings of the 43rd International ACM SIGIR conference on research and development in Information Retrieval . 639–648

2020
[7]

Xiangnan He, Lizi Liao, Hanwang Zhang, Liqiang Nie, Xia Hu, and Tat-Seng Chua. 2017. Neural collaborative filtering. In Proceedings of the 26th international conference on world wide web . 173–182

2017
[8]

Jonathan Ho, Ajay Jain, and Pieter Abbeel. 2020. Denoising diffusion probabilistic models. Advances in neural information processing systems 33 (2020), 6840–6851

2020

Show all 30 references

[9]

Yifan Hu, Yehuda Koren, and Chris Volinsky. 2008. Collaborative filtering for implicit feedback datasets. In 2008 Eighth IEEE international conference on data mining. Ieee, 263–272

2008
[10]

Yehuda Koren, Robert Bell, and Chris Volinsky. 2009. Matrix factorization tech- niques for recommender systems. Computer 42, 8 (2009), 30–37

2009
[11]

Zihao Li, Aixin Sun, and Chenliang Li. 2023. Diffurec: A diffusion model for sequential recommendation. ACM Transactions on Information Systems 42, 3 (2023), 1–28

2023
[12]

Dawen Liang, Rahul G Krishnan, Matthew D Hoffman, and Tony Jebara. 2018. Variational autoencoders for collaborative filtering. In Proceedings of the 2018 world wide web conference . 689–698

2018
[13]

Jinxin Liu, Yingyuan Xiao, Ke Zhu, Wenguang Zheng, and Ching-Hsien Hsu
[14]

Qidong Liu, Fan Yan, Xiangyu Zhao, Zhaocheng Du, Huifeng Guo, Ruiming Tang, and Feng Tian. 2023. Diffusion augmentation for sequential recommendation. In Proceedings of the 32nd ACM International Conference on Information and Knowledge Management. 1576–1586

2023
[15]

Calvin Luo. 2022. Understanding diffusion models: A unified perspective. arXiv preprint arXiv:2208.11970 (2022)

2022 arXiv
[16]

Steffen Rendle, Christoph Freudenthaler, Zeno Gantner, and Lars Schmidt-Thieme
[17]

Francesco Ricci, Lior Rokach, and Bracha Shapira. 2010. Introduction to rec- ommender systems handbook. In Recommender systems handbook . Springer, 1–35

2010
[18]

Harald Steck. 2019. Embarrassingly shallow autoencoders for sparse data. In The World Wide Web Conference. 3251–3257

2019
[19]

Jun Wang, Lantao Yu, Weinan Zhang, Yu Gong, Yinghui Xu, Benyou Wang, Peng Zhang, and Dell Zhang. 2017. Irgan: A minimax game for unifying generative and discriminative information retrieval models. In Proceedings of the 40th In- ternational ACM SIGIR conference on Research and...

2017
[20]

Wenjie Wang, Yiyan Xu, Fuli Feng, Xinyu Lin, Xiangnan He, and Tat-Seng Chua
[21]

Xiang Wang, Xiangnan He, Meng Wang, Fuli Feng, and Tat-Seng Chua. 2019. Neural graph collaborative filtering. In Proceedings of the 42nd international ACM SIGIR conference on Research and development in Information Retrieval . 165–174

2019
[22]

Yao Wu, Christopher DuBois, Alice X Zheng, and Martin Ester. 2016. Collabora- tive denoising auto-encoders for top-n recommender systems. In Proceedings of the ninth ACM international conference on web search and data mining . 153–162

2016
[23]

Zihao Wu, Xin Wang, Hong Chen, Kaidong Li, Yi Han, Lifeng Sun, and Wenwu Zhu. 2023. Diff4Rec: Sequential Recommendation with Curriculum-scheduled Diffusion Augmentation. In Proceedings of the 31st ACM International Conference on Multimedia. 9329–9335

2023
[24]

Kun Xu, Chongxuan Li, Jun Zhu, and Bo Zhang. 2020. Understanding and stabi- lizing GANs’ training dynamics using control theory. In International conference on machine learning. PMLR, 10566–10575

2020
[25]

Zhengyi Yang, Jiancan Wu, Zhicai Wang, Xiang Wang, Yancheng Yuan, and Xiangnan He. 2024. Generate What You Prefer: Reshaping Sequential Recommen- dation via Guided Diffusion. Advances in Neural Information Processing Systems 36 (2024)

2024
[26]

Yaochen Zhu and Zhenzhong Chen. 2022. Mutually-regularized dual collaborative variational auto-encoder for recommendation systems. InProceedings of The ACM Web Conference 2022. 2379–2387

2022
[27]

Yaochen Zhu, Liang Wu, Qi Guo, Liangjie Hong, and Jundong Li. 2024. Collab- orative large language model for recommender systems. In Proceedings of the ACM Web Conference 2024. 3162–3172

2024
[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
[2022]

In 2022 IEEE 25th International Conference on Computer Supported Cooperative Work in Design (CSCWD)

Hybrid Variational Autoencoder for Collaborative Filtering. In 2022 IEEE 25th International Conference on Computer Supported Cooperative Work in Design (CSCWD). IEEE, 251–256

2022
[2023]

In Proceedings of the 46th International ACM SIGIR Conference on Research and Development in Information Retrieval

Diffusion recommender model. In Proceedings of the 46th International ACM SIGIR Conference on Research and Development in Information Retrieval . 832–841

Pith tools

Reviewed August 9, 2026 · model on record in the stance chip above.

[1] [1]

Tommaso Carraro, Mirko Polato, Luca Bergamin, and Fabio Aiolli. 2022. Condi- tioned variational autoencoder for top-n item recommendation. In International Conference on Artificial Neural Networks . Springer, 785–796

work page 2022

[2] [2]

Dong-Kyu Chae, Jin-Soo Kang, Sang-Wook Kim, and Jung-Tae Lee. 2018. CFGAN: A generic collaborative filtering framework based on generative adversarial networks. In Proceedings of the 27th ACM international conference on information and knowledge management. 137–146

work page 2018

[3] [3]

Hanwen Du, Huanhuan Yuan, Zhen Huang, Pengpeng Zhao, and Xiaofang Zhou. 2023. Sequential recommendation with diffusion models. arXiv preprint arXiv:2304.04541 (2023)

arXiv 2023

[4] [4]

Gwangseok Han, Wonbin Kweon, Minsoo Kim, and Hwanjo Yu. 2024. Controlling Diversity at Inference: Guiding Diffusion Recommender Models with Targeted Category Preferences. arXiv preprint arXiv:2411.11240 (2024)

work page Pith review arXiv 2024

[5] [5]

GM Harshvardhan, Mahendra Kumar Gourisaria, Manjusha Pandey, and Sid- dharth Swarup Rautaray. 2020. A comprehensive survey and analysis of genera- tive models in machine learning. Computer Science Review 38 (2020), 100285

work page 2020

[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 Proceedings of the 43rd International ACM SIGIR conference on research and development in Information Retrieval . 639–648

2020

[7] [7]

Xiangnan He, Lizi Liao, Hanwang Zhang, Liqiang Nie, Xia Hu, and Tat-Seng Chua. 2017. Neural collaborative filtering. In Proceedings of the 26th international conference on world wide web . 173–182

2017

[8] [8]

Jonathan Ho, Ajay Jain, and Pieter Abbeel. 2020. Denoising diffusion probabilistic models. Advances in neural information processing systems 33 (2020), 6840–6851

2020