Data Context Adaptation for Accurate Recommendation with Additional Information

Bonhun Koo; Hyunsik Jeon; U Kang

REVIEW 3 major objections 4 minor 36 references

Data Context Adaptation for Accurate Recommendation with Additional Information

T0 review · 3 major / 4 minor · reviewed 2026-08-14 · deepseek-v4-flash

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

Pith's one-line read DaConA claims the best rating accuracy, up to 5.2% lower RMSE than the second-best method, by learning a separate linear projection for each data context before combining shared latent factors.

desk verdict A clean, well-specified deep CMF variant with a plausible architecture and reproducible code; the 'state-of-the-art' claim needs error bars or significance tests before it is fully convincing. read the letter →

arxiv 1908.08469 v1 pith:M6VTG2UH submitted 2019-08-22 cs.LG stat.ML

Hyunsik Jeon , Bonhun Koo , U Kang This is my paper · ORCID

classification cs.LGstat.ML

keywords collaborativefilteringmatrixfactorizationcollectivedatacontextadaptationneuralnetworkratingpredictionauxiliaryinformationrecommendersystems

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

The paper argues that collective recommendation methods fail when they force a single latent vector to play the same role in both the rating matrix and an auxiliary matrix, because the two matrices live in different data contexts. It proposes DaConA, which projects shared latent interaction vectors into each context using learned matrices, represents each entity with a multi-dimensional independence vector as well as an interaction vector, and scores pairs with a neural network instead of an inner product. If correct, DaConA achieves the lowest RMSE on six real-world datasets and subsumes MF, biased-MF, CMF, and biased-CMF as special cases.

What carries the argument

The data context adaptation layer: two learnable projection matrices, $D_X$ for the rating context and $D_Y$ for the auxiliary context, that map shared user, item, and auxiliary-entity interaction vectors into context-appropriate representations. These projected vectors are combined by element-wise product, concatenated with latent independence vectors, and fed into fully-connected networks with Tanh activations, which lets the model capture non-linear interactions.

What would settle it

Construct a synthetic dataset with known ground truth where the rating process uses per-entity rotations between contexts, e.g., $r_{ij} = g((W_i u_i) \circ (W_j v_j))$ with different $W$ per entity. If DaConA's single-shared-projection model cannot beat a per-entity projection baseline or DaConA-withoutCA on this dataset, the shared linear projection assumption is falsified.

Watch

Extended reading notes

Core claim

The central claim is that a data context adaptation layer resolves the context mismatch in collective matrix factorization. DaConA learns $D_X$ and $D_Y$, two linear projections that map shared latent interaction vectors into the rating context and the auxiliary context, then computes element-wise products of the projected vectors, concatenates them with per-context independence vectors, and predicts ratings and auxiliary entries through two multilayer perceptrons. The paper further shows that setting the projections to identities, removing independence vectors, and using a single linear layer reduces DaConA exactly to MF and CMF, so the new method is a strict generalization with richer modeling capacity.

Load-bearing premise

The method assumes that a single shared linear projection per data context is enough to translate every user, item, and auxiliary entity's interaction vector into that context; if the needed translation differs sharply per entity or is non-linear, the core adaptation idea loses its justification.

Editorial extensions

If this is right

If DaConA is correct, collective recommendation methods can treat data context as a learned linear map in latent space rather than retraining separate embeddings per matrix.
Replacing scalar bias terms with vector-valued independence factors increases modeling capacity for user and item idiosyncrasies.
Non-linear scoring through neural networks outperforms inner-product scoring on both user-coupled and item-coupled auxiliary data.
Because DaConA generalizes MF, biased-MF, CMF, and biased-CMF, any improvement it demonstrates transfers to those classical frameworks as special cases.
The architecture extends to multiple auxiliary matrices, such as combining item-genre and user-trust information simultaneously.

Reading between the lines

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

The shared-per-context projection is a testable symmetry assumption: in datasets where users and items occupy very different latent subspaces, per-entity-type projections might outperform the single shared $D_X$, an experiment the paper does not run.
A natural extension the paper leaves implicit is cold-start recommendation, where a new entity's latent interaction vector could be initialized from its auxiliary features and immediately projected into the rating context.
The data context adaptation idea is not specific to ratings: it applies to any paired relational matrices, such as document-word and author-venue, wherever the same entities appear in two semantically different relations.

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

You should know two things about this paper. First, it is a genuinely clean method paper: DaConA combines data context adaptation (a linear projection per matrix), vector-valued independence terms, and neural scoring into one collective matrix factorization framework, and the special-case reductions to MF/CMF are correct by construction. The source code and datasets are available, which is more than most papers at this level. Second, the headline accuracy claim is not yet established: Table IV reports a single RMSE per method/dataset with no error bars or significance tests, and the improvements over the second-best are as small as 1.5% on ML-1m. That gap could easily be within run-to-run or split-to-split noise. The stress-test note has it right.

What is actually new is the specific architecture, especially the shared per-context projection matrix and the vector-valued latent independence terms. Prior work uses projections per model (SREPS) or scalar biases, but not a single shared projection per data context with non-linear scoring on top. The ablation in Figure 4 comparing DaConA to DaConA-sep (separate projections per entity) is a sensible way to justify the sharing decision, and the result supports it.

Where it is soft: beyond the missing error bars, the ablations reuse the full model's optimized hyperparameters, which can tilt things in favor of the full model. That is common in this literature, but it is still a bias worth noting. The generalization proof is really a 'contains as special cases' argument, which is fine, but it is not a theorem about approximation capacity. The assumption that a single linear projection suffices for each context is empirically defended, not theoretically.

The math itself is not circular and the derivations are straightforward. No invented entities, no prediction defined from fitted parameters.

Who gets value: anyone working on side-information recommendation who wants a solid, reproducible baseline that extends CMF to non-linear scoring. It is not a breakaway idea, but it is a useful one.

My recommendation: send it to peer review. The empirical claim needs variance estimates or significance tests and fairer ablations before it can be called state-of-the-art, but the architecture and the released code justify referee time.

Referee Report

3 major / 4 minor

Summary. The paper proposes DaConA, a neural-network method for rating prediction using both a rating matrix and an auxiliary matrix (user-coupled or item-coupled). The architecture has three main ingredients: (i) a data context adaptation layer that projects shared latent interaction vectors into the rating context and the auxiliary context with per-context linear matrices D_X and D_Y; (ii) separate latent interaction and latent independence vectors per entity, with the independence vectors not restricted to scalars; and (iii) non-linear scoring via fully-connected networks f_X and f_Y on the element-wise product of adapted interaction vectors and the concatenated independence vectors. The paper shows that DaConA reduces to MF, biased-MF, CMF, and biased-CMF under specific parameter settings, and reports RMSE/MAE comparisons on six real-world datasets, claiming state-of-the-art accuracy with up to 5.2% lower RMSE than the second-best method.

Significance. If the performance claims hold, the paper makes a useful and clearly explained contribution: the context-adaptation layer is a simple yet plausible extension of collective matrix factorization, the independence-vector generalization is more expressive than the scalar biases of biased-MF/biased-CMF, and the non-linear scoring function is a natural fit with modern recommender models. The paper also deserves credit for giving explicit special-case reductions to MF and CMF, which are correct by construction, and for providing a source-code/dataset link. However, the significance is conditional: the headline accuracy claim rests entirely on single-point comparisons without variance or significance testing, so the empirical advantage over strong baselines is not yet established at the level claimed.

major comments (3)

[§IV-B, Table IV and Fig. 1] The central claim that DaConA achieves state-of-the-art accuracy rests entirely on single-point RMSE/MAE comparisons. Table IV reports one value per method and dataset with no standard deviations, no repeated data splits, and no significance tests; the reported gains over the second-best method range from 1.5% to 5.2% in RMSE, which may be within run-to-run or split-to-split noise. This is not a hypothetical concern: Section IV-C reports results averaged over ten runs, so the authors have the machinery to produce variance estimates, and the main table should include them. Please report standard deviations (or confidence intervals) and significance tests for the differences in Table IV, or alternatively soften the state-of-the-art claim to a conditional one.
[§III-F, Algorithm 1 and Eq. (1)] The stated objective is the weighted sum L = (1−α)lossX + α lossY, but Algorithm 1 does not minimize this objective. In lines 7–8 the shared parameter block V is updated using only (1−α)lossX, and in lines 14–15 using only α lossY, so no update step ever forms the gradient of the combined loss with respect to V. This matters because V appears in both lossX and lossY. Please either provide an optimization-theoretic justification for the alternating procedure as a solver for Eq. (1), or revise the exposition to state that the algorithm minimizes the two terms in an alternating fashion rather than the weighted sum.
[§IV-D, Figure 4] The ablation comparison between DaConA, DaConA-withoutCA, and DaConA-sep sets all hyperparameters to the optimal settings of the full DaConA model. If the ablation variants have different optimal hyperparameters (for instance, a different ds or α), the reported differences in Figure 4 may partly reflect mistuned ablations rather than the effect of the context-adaptation layer. Please re-tune the ablations independently and report their optimal hyperparameters, or justify why the full model's settings are appropriate for all variants.

minor comments (4)

[§IV-C, Figure 3] In the dimension sweep, dc is set to 40−2ds, so varying ds changes both the independence capacity and the interaction capacity simultaneously; the conclusion about balancing the two factors should be stated with this coupling in mind.
[Table V] The hyperparameter table is difficult to parse because per-dataset values are packed into single rows; please format it as a method-by-dataset table or add explicit column headers so that each value can be traced to its dataset.
[§III-G] For the MF and CMF special-case reductions, please state explicitly that the one-layer network is a sum layer, so that the element-wise product followed by summation equals the inner product; the phrase 'all weights of the layer to 1' is ambiguous about whether it also sums over the element-wise product dimension.
[Figure 1] The caption uses 'N/A' without explaining what it refers to; the text says FM is omitted, but the N/A markers appear to indicate that SREPS is unavailable for the item-coupled datasets. Please clarify the caption.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: DaConA's accuracy claims rest on held-out test ratings, and its MF/CMF special-case results are explicit parameter restrictions, not fitted inputs presented as predictions.

full rationale

The paper's derivation chain is self-contained. The predictive model in Eqs. (5)-(6) defines predictions from latent interaction vectors, latent independence vectors, and learned context projection matrices, and the objective in Eqs. (1)-(3) is minimized on observed entries only, with test ratings held out (80/20 split, Section IV-A). The central accuracy claim (Table IV, Figure 1) is an empirical comparison against external baselines on held-out ratings, so no prediction is defined in terms of fitted parameters. The generalization claims in Section III-G are explicit parameter choices (alpha=0, identity projection, ds=0, unit weights, no activation) that reduce DaConA to MF, Biased-MF, CMF, and Biased-CMF; these are mathematical special-case reductions, not evidence for predictive performance, and thus not circular. The effectiveness of the data context adaptation layer is tested by an independent ablation (DaConA-withoutCA, DaConA-sep, Section IV-D) under fixed hyperparameters, and the paper reports that sharing the projection matrix per context improves results. Citations are to external prior work (e.g., NeuMF, CMF, SREPS, Hybrid-CDL); there is no load-bearing self-citation chain or imported uniqueness theorem. The absence of error bars in the main comparison is a statistical-reporting concern, not a circularity concern.

Assumptions & free parameters 5 free parameters · 5 assumptions · 0 invented entities

The central claims rest on standard supervised-learning assumptions and a small set of hyperparameters tuned on validation. No new physical entities or exotic axioms are introduced.

free parameters (5)

alpha (context balance) = 0.4 / 0.8 / 0.9 / 0.2 / 0.9 / 0.6
Balances rating loss vs auxiliary loss; grid-searched per dataset in Table V.
lambda (L2 regularization) = 1e-5 for all DaConA runs
Regularization strength; grid-searched per dataset in Table V.
ds (independence vector dimension) = 13 / 14 / 11 / 7 / 10 / 11
Controls capacity for independence factors; tuned per dataset in Table V.
learning rate = 1e-3 or 1e-4 depending on dataset
Adam step size; grid-searched per dataset in Table V.
network depth = [40, 20, 10] hidden units
Tower structure chosen after experiments; depth 3 works best across datasets.

assumptions (5)

domain assumption Shared latent interaction vectors U, V, C can represent both rating and auxiliary contexts after multiplication by a per-context projection matrix.
Used in Eq. (5) where D_X and D_Y map the same vectors into two contexts.
domain assumption Entries of the auxiliary matrix Y (trust, genre) can be modeled as real-valued targets with squared error.
lossY in Eq. (3) applies an L2 loss to Y entries, even though trust and genre entries are often binary.
domain assumption A random 80/20 split of ratings approximates the deployment setting for measuring recommendation accuracy.
Section IV-A creates one random split per dataset; no temporal or user-based split is tested.
standard math Alternating optimization with Adam converges to a solution good enough for comparing methods.
Algorithm 1 alternately updates blocks; no convergence guarantee is provided.
domain assumption The element-wise product of projected interaction vectors plus concatenated independence vectors is a sufficient input representation for the scoring network.
Eq. (6) fixes this interaction form without comparing to other fusion functions such as concatenation only or bilinear layers.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Data Context Adaptation for Accurate Recommendation with Additional Information." pith.science (2026). https://pith.science/paper/M6VTG2UH

@misc{pith2026190808469,
  author       = {Pith},
  title        = {Pith review of: Data Context Adaptation for Accurate Recommendation with Additional Information},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/M6VTG2UH}},
  note         = {Machine review of arXiv:1908.08469}
}

read the original abstract

Given a sparse rating matrix and an auxiliary matrix of users or items, how can we accurately predict missing ratings considering different data contexts of entities? Many previous studies proved that utilizing the additional information with rating data is helpful to improve the performance. However, existing methods are limited in that 1) they ignore the fact that data contexts of rating and auxiliary matrices are different, 2) they have restricted capability of expressing independence information of users or items, and 3) they assume the relation between a user and an item is linear. We propose DaConA, a neural network based method for recommendation with a rating matrix and an auxiliary matrix. DaConA is designed with the following three main ideas. First, we propose a data context adaptation layer to extract pertinent features for different data contexts. Second, DaConA represents each entity with latent interaction vector and latent independence vector. Unlike previous methods, both of the two vectors are not limited in size. Lastly, while previous matrix factorization based methods predict missing values through the inner-product of latent vectors, DaConA learns a non-linear function of them via a neural network. We show that DaConA is a generalized algorithm including the standard matrix factorization and the collective matrix factorization as special cases. Through comprehensive experiments on real-world datasets, we show that DaConA provides the state-of-the-art accuracy.

Figures

Figures reproduced from arXiv: 1908.08469 by the authors.

**Figure 1.** DACONA shows the best accuracy (lowest RMSE) for six real-world datasets; each percentage indicates the error reduction rate compared to the second best one in each dataset. We omit FM in the plot since it shows far worse performance than the other methods. SREPS is available only for usercoupled datasets. most popular method for data context-aware recommendations. CMF comprises two tied MF models which share a sin… view at source ↗

**Figure 2.** Architecture of DACONA for item coupled dataset. respectively. We then formulate the objective function (to be minimized) of DACONA as follows: L = (1 − α)lossX + αlossY , (1) where lossX = 1 2 X (i,j)∈ΩX (Xˆ ij − Xij ) 2 + λ 2 RegX, (2) lossY = 1 2 X (j,k)∈ΩY (Yˆ jk − Yjk) 2 + λ 2 RegY . (3) ΩX and ΩY are sets of indices of observable entries in X and Y , respectively. Xˆ ij and Yˆ jk are predicted entries from DAC… view at source ↗

**Figure 4.** DACONA outperforms DACONA-withoutCA, showing that data context adaptation improves the accuracy. DACONA also outperforms DACONA-sep, showing that sharing the same data context adaptation matrix improves the accuracy. that more performance improvements are observed for larger and sparser data (Epinions, Ciao-u, and Ciao-i). Despite adding the learnable matrices, DACONA-sep shows little change in performance when co… view at source ↗

Figures from the paper (2 more)

**Figure 5.** Figure 5: Using 3 layers instead of too small or too large layers [PITH_FULL_IMAGE:figures/full_fig_p009_5.png]

**Figure 7.** Figure 7: DACONA gives the best accuracy even when using two auxiliary matrices. Each arrow indicates the reduction rate of DACONA’s RMSE compared to hybrid-CDL’s RMSE. where fZ is a fully-connected neural network. The predictive models for X and Y are defined in Equation (6) (d…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

36 extracted references · 35 canonical work pages

[1]

Matrix factorization techniques for recommender systems,

Y . Koren, R. M. Bell, and C. V olinsky, “Matrix factorization techniques for recommender systems,” IEEE Computer, 2009

work page 2009
[2]

Probabilistic matrix factorization,

R. Salakhutdinov and A. Mnih, “Probabilistic matrix factorization,” in NIPS, 2007

work page 2007
[3]

Neural network matrix factorization,

G. K. Dziugaite and D. M. Roy, “Neural network matrix factorization,” CoRR, 2015

work page 2015
[4]

Neural collaborative ﬁltering,

X. He, L. Liao, H. Zhang, L. Nie, X. Hu, and T. Chua, “Neural collaborative ﬁltering,” in WWW, 2017

work page 2017
[5]

Relational learning via collective matrix factorization,

A. P. Singh and G. J. Gordon, “Relational learning via collective matrix factorization,” in SIGKDD, 2008

work page 2008
[6]

Factorization machines,

S. Rendle, “Factorization machines,” in ICDM, 2010

work page 2010
[7]

Factorization machines with libfm,

S. Rendle, “Factorization machines with libfm,” ACM TIST, 2012

work page 2012
[8]

Higher-order factor- ization machines,

M. Blondel, A. Fujino, N. Ueda, and M. Ishihata, “Higher-order factor- ization machines,” in NIPS, 2016

work page 2016

Show all 36 references

[9]

Social recommendation with an essential preference space,

C. Liu, C. Zhou, J. Wu, Y . Hu, and L. Guo, “Social recommendation with an essential preference space,” in AAAI, 2018

2018
[10]

A hybrid collab- orative ﬁltering model with deep structure for recommender systems,

X. Dong, L. Yu, Z. Wu, Y . Sun, L. Yuan, and F. Zhang, “A hybrid collab- orative ﬁltering model with deep structure for recommender systems,” in AAAI, 2017

2017
[11]

Sorec: social recommendation using probabilistic matrix factorization,

H. Ma, H. Yang, M. R. Lyu, and I. King, “Sorec: social recommendation using probabilistic matrix factorization,” in CIKM, 2008

2008
[12]

Learning to recommend with social trust ensemble,

H. Ma, I. King, and M. R. Lyu, “Learning to recommend with social trust ensemble,” in SIGIR, 2009

2009
[13]

A matrix factorization technique with trust propagation for recommendation in social networks,

M. Jamali and M. Ester, “A matrix factorization technique with trust propagation for recommendation in social networks,” in RecSys, 2010

2010
[14]

Recommender systems with social regularization,

H. Ma, D. Zhou, C. Liu, M. R. Lyu, and I. King, “Recommender systems with social regularization,” in WSDM, 2011

2011
[15]

Social collaborative ﬁltering by trust,

B. Yang, Y . Lei, D. Liu, and J. Liu, “Social collaborative ﬁltering by trust,” in IJCAI, 2013

2013
[16]

Recommendation with social dimensions,

J. Tang, S. Wang, X. Hu, D. Yin, Y . Bi, Y . Chang, and H. Liu, “Recommendation with social dimensions,” in AAAI, 2016

2016
[17]

Deep collaborative ﬁltering via marginal- ized denoising auto-encoder,

S. Li, J. Kawale, and Y . Fu, “Deep collaborative ﬁltering via marginal- ized denoising auto-encoder,” in CIKM, 2015

2015
[18]

Collaborative topic modeling for recommend- ing scientiﬁc articles,

C. Wang and D. M. Blei, “Collaborative topic modeling for recommend- ing scientiﬁc articles,” in SIGKDD, 2011

2011
[19]

Collaborative deep learning for recommender systems,

H. Wang, N. Wang, and D. Yeung, “Collaborative deep learning for recommender systems,” in SIGKDD, 2015

2015
[20]

Convolutional matrix factorization for document context-aware recommendation,

D. H. Kim, C. Park, J. Oh, S. Lee, and H. Yu, “Convolutional matrix factorization for document context-aware recommendation,” in RecSys, 2016

2016
[21]

A survey of collaborative ﬁltering techniques,

X. Su and T. M. Khoshgoftaar, “A survey of collaborative ﬁltering techniques,” Adv. Artiﬁcial Intellegence, 2009

2009
[22]

Understanding the difﬁculty of training deep feedforward neural networks,

X. Glorot and Y . Bengio, “Understanding the difﬁculty of training deep feedforward neural networks,” in AISTATS, 2010

2010
[23]

Adam: A method for stochastic optimization,

D. P. Kingma and J. Ba, “Adam: A method for stochastic optimization,” in ICLR, 2015

2015
[24]

Deep neural networks for youtube recommendations,

P. Covington, J. Adams, and E. Sargin, “Deep neural networks for youtube recommendations,” in RecSys, 2016

2016
[25]

A probabilistic model for using social networks in personalized item recommendation,

A. J. Chaney, D. M. Blei, and T. Eliassi-Rad, “A probabilistic model for using social networks in personalized item recommendation,” in RecSys, 2015

2015
[26]

Trustsvd: Collaborative ﬁltering with both the explicit and implicit inﬂuence of user trust and of item ratings,

G. Guo, J. Zhang, and N. Yorke-Smith, “Trustsvd: Collaborative ﬁltering with both the explicit and implicit inﬂuence of user trust and of item ratings,” in AAAI, 2015

2015
[27]

ETAF: an extended trust antecedents framework for trust prediction,

G. Guo, J. Zhang, D. Thalmann, and N. Yorke-Smith, “ETAF: an extended trust antecedents framework for trust prediction,” in ASONAM, 2014

2014
[28]

Circle-based recommendation in online social networks,

X. Yang, H. Steck, and Y . Liu, “Circle-based recommendation in online social networks,” in SIGKDD, 2012

2012
[29]

An experimental study on implicit social recommendation,

H. Ma, “An experimental study on implicit social recommendation,” in SIGIR, 2013

2013
[30]

Dynamic connection-based social group recommendation,

D. Qin, X. Zhou, L. Chen, G. Huang, and Y . Zhang, “Dynamic connection-based social group recommendation,” IEEE TKDE, 2018

2018
[31]

Integrating collaborative ﬁltering and sentiment analysis: A rating inference approach,

C. W. Leung, S. C. Chan, and F.-l. Chung, “Integrating collaborative ﬁltering and sentiment analysis: A rating inference approach,” in Pro- ceedings of the ECAI 2006 workshop on recommender systems , 2006

2006
[32]

Integrating reviews into personalized ranking for cold start recommendation,

G. Hu and X. Dai, “Integrating reviews into personalized ranking for cold start recommendation,” in PAKDD, 2017

2017
[33]

Aspect based recommendations: Recommending items with the most valuable aspects based on user reviews,

K. Bauman, B. Liu, and A. Tuzhilin, “Aspect based recommendations: Recommending items with the most valuable aspects based on user reviews,” in SIGKDD, 2017

2017
[34]

Recommender systems based on user reviews: the state of the art,

L. Chen, G. Chen, and F. Wang, “Recommender systems based on user reviews: the state of the art,” User Model. User-Adapt. Interact. , 2015

2015
[35]

Stacked denoising autoencoders: Learning useful representations in a deep network with a local denoising criterion,

P. Vincent, H. Larochelle, I. Lajoie, Y . Bengio, and P. Manzagol, “Stacked denoising autoencoders: Learning useful representations in a deep network with a local denoising criterion,” JMLR, 2010

2010
[36]

Deep hybrid recommender systems via exploiting document context and statistics of items,

D. H. Kim, C. Park, J. Oh, and H. Yu, “Deep hybrid recommender systems via exploiting document context and statistics of items,” Inf. Sci., 2017. APPENDIX A. Using Multiple Auxiliary Information Extension. DACONA supports using multiple auxiliary ma- trices. Assume we are give...

2017

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.

[1] [1]

Matrix factorization techniques for recommender systems,

Y . Koren, R. M. Bell, and C. V olinsky, “Matrix factorization techniques for recommender systems,” IEEE Computer, 2009

work page 2009

[2] [2]

Probabilistic matrix factorization,

R. Salakhutdinov and A. Mnih, “Probabilistic matrix factorization,” in NIPS, 2007

work page 2007

[3] [3]

Neural network matrix factorization,

G. K. Dziugaite and D. M. Roy, “Neural network matrix factorization,” CoRR, 2015

work page 2015

[4] [4]

Neural collaborative ﬁltering,

X. He, L. Liao, H. Zhang, L. Nie, X. Hu, and T. Chua, “Neural collaborative ﬁltering,” in WWW, 2017

work page 2017

[5] [5]

Relational learning via collective matrix factorization,

A. P. Singh and G. J. Gordon, “Relational learning via collective matrix factorization,” in SIGKDD, 2008

work page 2008

[6] [6]

Factorization machines,

S. Rendle, “Factorization machines,” in ICDM, 2010

work page 2010

[7] [7]

Factorization machines with libfm,

S. Rendle, “Factorization machines with libfm,” ACM TIST, 2012

work page 2012

[8] [8]

Higher-order factor- ization machines,

M. Blondel, A. Fujino, N. Ueda, and M. Ishihata, “Higher-order factor- ization machines,” in NIPS, 2016

work page 2016