"Collaborative Deep Learning for Recommender Systems"의 두 판 사이의 차이

ph
이동: 둘러보기, 검색
잔글
잔글 (→‎CDL)
47번째 줄: 47번째 줄:
  
 
====Prediction====
 
====Prediction====
 +
 +
===References===
  
  

2017년 6월 21일 (수) 16:47 판

[1409.2944] Collaborative Deep Learning for Recommender Systems
by H Wang - ‎2014 - ‎Cited by 113 - ‎Related articles
https://arxiv.org/abs/1409.2944

code?

https://github.com/akash13singh/mxnet-for-cdl

intro

  • CF는 sparse할 때 약하다. CTR(Collaborative topic regression)이 등장.
    • CF cannot be used for new products
  • hierarchical Bayesian model인 CDL(Collaborative Deep Learning)을 제안한다.
    • deep representation learning은 dl로,
    • rating은 CF로.
  • CF의 단점때문에 보조정보(auxiliary information)을 이용하는 method 많이들씀
    1. loosely coupled
      • 보조정보 처리 후 CF에 넣음
    2. tightly coupled
      • the rating information can guide the learning of features
      • Collaborative topic regression (CTR) [1]
        = Latent Dirichlet allocation (LDA) [2](as a topic model) + probabilistic matrix factorization (PMF) [3](as a model-based CF).
      • auxiliary info도 sparse하면 잘 안된다.
  • We first present a Bayesian formulation of a deep learning model called stacked denoising autoencoder (SDAE) [4]. With this, we then present our CDL model which tightly couples deep representation learning for the content information and collaborative filtering for the ratings (feedback) matrix, allowing two-way interaction between the two.
    • SDAE말고, RBM, CNN, RNN같은거 써도 된다.
  • contributions
    • Unlike previous deep learning models which use simple target like classification [5]and reconstruction [6], we propose to use CF as a more complex target in a probabilistic framework.
    • Besides the algorithm for attaining maximum a poste- riori (MAP) estimates, we also derive a sampling-based algorithm for the Bayesian treatment of CDL, which, interestingly, turns out to be a Bayesian generalized version of back-propagation.
    • etc.

notation

  • \(\mathbf{X}_c\)는 \(J\times S\) matrix. J items가 있고, 각 item이 S dimension. c는 clean을 뜻함. SDAE의 input이 된다(noise corrupted matrix는 \(\mathbf{X}_0\))
  • \(\mathbf{R}\)은 \(I\times J\) matrix. User수가 \(I\).
  • Note that an L/2-layer SDAE corresponds to an L-layer network.

CDL

Stacked Denoising Autoencoders

AE인데 그냥 stack한것.

Generalized Bayesian SDAE

Note that while generation of the clean input \(\mathbf{X}_c\) from \(\mathbf{X}_L\) is part of the generative process of the Bayesian SDAE, generation of the noise-corrupted input \(\mathbf{X}_0\) from \(\mathbf{X}_c\) is an artificial noise injection process to help the SDAE learn a more robust feature representation.

Collaborative Deep Learning

Maximum A Posteriori Estimates

Prediction

References

  1. ↑ C. Wang and D. M. Blei. Collaborative Topic Modeling for Recommending Scientific Articles. In KDD, pages 448–456, 2011.
  2. ↑ D. M. Blei, A. Y. Ng, and M. I. Jordan. Latent Dirichlet allocation. JMLR, 3:993–1022, 2003.
  3. ↑ R. Salakhutdinov and A. Mnih. Probabilistic matrix factorization. In NIPS, pages 1257–1264, 2007.
  4. ↑ P. Vincent, H. Larochelle, I. Lajoie, Y. Bengio, and P.-A. Manzagol. Stacked denoising autoencoders: Learning useful representations in a deep network with a local denoising criterion. JMLR, 11:3371–3408, 2010.
  5. ↑ N. Kalchbrenner, E. Grefenstette, and P. Blunsom. A convolutional neural network for modelling sentences. ACL, pages 655–665, 2014.
  6. ↑ P. Vincent, H. Larochelle, I. Lajoie, Y. Bengio, and P.-A. Manzagol. Stacked denoising autoencoders: Learning useful representations in a deep network with a local denoising criterion. JMLR, 11:3371–3408, 2010.

blog comments powered by Disqus

원본 주소 "http://samediff.kr/wiki/index.php?title=Collaborative_Deep_Learning_for_Recommender_Systems&oldid=12234"