Collaborative Deep Learning for Recommender Systems - 편집 역사

Admin: /* Maximum A Posteriori Estimates */

2017-07-05T10:16:36Z

Maximum A Posteriori Estimates

Admin: /* Maximum A Posteriori Estimates */

2017-07-05T10:14:20Z

Maximum A Posteriori Estimates

Admin: /* Maximum A Posteriori Estimates */

2017-07-05T10:13:41Z

Maximum A Posteriori Estimates

Admin: /* intro */

2017-07-05T09:33:01Z

intro

Admin: /* Generalized Bayesian SDAE */

2017-07-05T06:29:34Z

Generalized Bayesian SDAE

Admin: /* REFERENCES */

2017-07-05T06:28:57Z

REFERENCES

Admin: /* intro */

2017-06-26T06:17:09Z

intro

Admin: /* code */

2017-06-21T16:51:41Z

code

Admin: /* code? */

2017-06-21T09:43:24Z

code?

Admin: /* code? */

2017-06-21T09:37:02Z

code?

@@ 61번째 줄: / 61번째 줄: @@
 update rule(논문참조)에 따라 iterative하게 update한다.
-단, [[Collaborative Topic Modeling for Recommending Scientific Articles]]의 경우 \(v_j\)를 수정하는 \(\theta\)가  미리 계산될 수 있어서, \(\mathbf{U}, \mathbf{V}\)만 update하지만, 이 경우는 neural net학습과 동시에 진행되기 때문에 \(\mathbf{U}, \mathbf{V}, \mathbf{W}\)순서로 학습된다.
+단, [[Collaborative Topic Modeling for Recommending Scientific Articles]]의 경우 \(v_j\)를 수정하는 \(\theta\)가  미리 계산될 수 있어서, \(\mathbf{U}, \mathbf{V}\)만 update하지만, 이 경우는 neural net학습과 동시에 진행되기 때문에 \(\mathbf{U}, \mathbf{V}, \mathbf{W}, \mathbf{b}\)순서로 iteration한다.
 ====Prediction====

@@ 61번째 줄: / 61번째 줄: @@
 update rule(논문참조)에 따라 iterative하게 update한다.
-단, [[Collaborative Topic Modeling for Recommending Scientific Articles]]의 경우 \(v_j\)를 수정하는 \(\theta\)가  미리 계산될 수 있어서, \(\mathbf{U}, \mathbf{V}\)만 update하지만, 이 경우는 neural net학습과 동시에 진행되기 때문에 (\mathbf{U}, \mathbf{V}, \mathbf{W}\)순서로 학습된다.
+단, [[Collaborative Topic Modeling for Recommending Scientific Articles]]의 경우 \(v_j\)를 수정하는 \(\theta\)가  미리 계산될 수 있어서, \(\mathbf{U}, \mathbf{V}\)만 update하지만, 이 경우는 neural net학습과 동시에 진행되기 때문에 \(\mathbf{U}, \mathbf{V}, \mathbf{W}\)순서로 학습된다.
 ====Prediction====

← 이전 판		2017년 7월 5일 (수) 10:13 판
59번째 줄:		59번째 줄:

	====Maximum A Posteriori Estimates====		====Maximum A Posteriori Estimates====
		+	update rule(논문참조)에 따라 iterative하게 update한다.
		+
		+	단, [[Collaborative Topic Modeling for Recommending Scientific Articles]]의 경우 \(v_j\)를 수정하는 \(\theta\)가 미리 계산될 수 있어서, \(\mathbf{U}, \mathbf{V}\)만 update하지만, 이 경우는 neural net학습과 동시에 진행되기 때문에 (\mathbf{U}, \mathbf{V}, \mathbf{W}\)순서로 학습된다.

	====Prediction====		====Prediction====

@@ 22번째 줄: / 22번째 줄: @@
 ===intro===
-* CF는 sparse할 때 약하다. CTR(Collaborative topic regression)이 등장.
+* CF는 sparse할 때 약하고 cold start problem이 있다.
-** CF cannot be used for new products
+** CTR(Collaborative topic regression)이 등장.
 * hierarchical Bayesian model인 CDL(Collaborative Deep Learning)을 제안한다.
 ** deep representation learning은 dl로,

@@ 54번째 줄: / 54번째 줄: @@
 ====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.
+<blockquote>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.</blockquote>
 ====Collaborative Deep Learning====

@@ 82번째 줄: / 82번째 줄: @@
 ===REFERENCES===
 <small>
-[1] D. Agarwal and B.-C. Chen. Regression-based latent factor models. In KDD, pages 19–28, 2009.<br>
+<del>[1] D. Agarwal and B.-C. Chen. Regression-based latent factor models. In KDD, pages 19–28, 2009.<br>
 [2] P. Baldi and P. J. Sadowski. Understanding dropout. In NIPS, pages 2814–2822, 2013.<br>
 [3] Y. Bengio, L. Yao, G. Alain, and P. Vincent. Generalized denoising auto-encoders as generative models. In NIPS, pages 899–907, 2013.<br>
@@ 93번째 줄: / 93번째 줄: @@
 [10] A. Graves, S. Ferna ́ndez, F. J. Gomez, and J. Schmidhuber. Connectionist temporal classification: labelling unsegmented sequence data with recurrent neural networks. In ICML, pages 369–376, 2006.<br>
 [11] G. E. Hinton, N. Srivastava, A. Krizhevsky, I. Sutskever, and R. Salakhutdinov. Improving neural networks by preventing co-adaptation of feature detectors. CoRR, abs/1207.0580, 2012.<br>
-[12] L.Hu,J.Cao,G.Xu,L.Cao,Z.Gu,andC.Zhu. Personalized recommendation via cross-domain triadic factorization. In WWW, pages 595–606, 2013.<br>
+[12] L.Hu,J.Cao,G.Xu,L.Cao,Z.Gu,andC.Zhu. Personalized recommendation via cross-domain triadic factorization. In WWW, pages 595–606, 2013.<br> </del>
 [13] Y. Hu, Y. Koren, and C. Volinsky. Collaborative filtering for implicit feedback datasets. In ICDM, pages 263–272, 2008.<br>
-[14] M. I. Jordan, Z. Ghahramani, T. Jaakkola, and L. K. Saul. An introduction to variational methods for graphical models. Machine Learning, 37(2):183–233, 1999.<br>
+<del>[14] M. I. Jordan, Z. Ghahramani, T. Jaakkola, and L. K. Saul. An introduction to variational methods for graphical models. Machine Learning, 37(2):183–233, 1999.<br>
 [15] N. Kalchbrenner, E. Grefenstette, and P. Blunsom. A convolutional neural network for modelling sentences. ACL, pages 655–665, 2014.<br>
 [16] A. Krizhevsky, I. Sutskever, and G. E. Hinton. ImageNet classification with deep convolutional neural networks. In NIPS, pages 1106–1114, 2012.<br>
@@ 114번째 줄: / 114번째 줄: @@
 [31] R. S. Strichartz. A Guide to Distribution Theory and Fourier Transforms. World Scientific, 2003.<br>
 [32] 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.<br>
-[33] S. Wager, S. Wang, and P. Liang. Dropout training as adaptive regularization. In NIPS, pages 351–359, 2013.<br>
+[33] S. Wager, S. Wang, and P. Liang. Dropout training as adaptive regularization. In NIPS, pages 351–359, 2013.<br> </del>
 [34] C. Wang and D. M. Blei. [[Collaborative Topic Modeling for Recommending Scientific Articles|Collaborative topic modeling for recommending scientific articles]]. In KDD, pages 448–456, 2011.<br>
-[35] H. Wang, B. Chen, and W.-J. Li. Collaborative topic regression with social regularization for tag recommendation. In IJCAI, pages 2719–2725, 2013.<br>
+<del>[35] H. Wang, B. Chen, and W.-J. Li. Collaborative topic regression with social regularization for tag recommendation. In IJCAI, pages 2719–2725, 2013.<br>
 [36] H. Wang and W. Li. Relational collaborative topic regression for recommender systems. TKDE, 27(5):1343–1355, 2015.<br>
 [37] H. Wang, X. Shi, and D. Yeung. Relational stacked denoising autoencoder for tag recommendation. In AAAI, pages 3052–3058, 2015.<br>
@@ 122번째 줄: / 122번째 줄: @@
 [39] X. Wang and Y. Wang. Improving content-based and hybrid music recommendation using deep learning. In ACM MM, pages 627–636, 2014.<br>
 [40] W. Zhang, H. Sun, X. Liu, and X. Guo. Temporal qos-aware web service recommendation via non-negative tensor factorization. In WWW, pages 585–596, 2014.<br>
-[41] K. Zhou and H. Zha. Learning binary codes for collaborative filtering. In KDD, pages 498–506, 2012.<br>
+[41] K. Zhou and H. Zha. Learning binary codes for collaborative filtering. In KDD, pages 498–506, 2012.<br> </del>
 </small>

@@ 41번째 줄: / 41번째 줄: @@
 * contributions
 ** Unlike previous deep learning models which use simple target like classification <ref>N. Kalchbrenner, E. Grefenstette, and P. Blunsom. A convolutional neural network for modelling sentences. ACL, pages 655–665, 2014.</ref>and reconstruction <ref>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.</ref>, 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.
+** Besides the algorithm for attaining maximum a posteriori (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.

← 이전 판		2017년 6월 21일 (수) 16:51 판
18번째 줄:		18번째 줄:

	기타 : https://github.com/akash13singh/mxnet-for-cdl		기타 : https://github.com/akash13singh/mxnet-for-cdl
		+
		+	코드 분석 및 실제 실험 : [[Collaborative Deep Learning for Recommender Systems : Practice]]

	===intro===		===intro===

@@ 4번째 줄: / 4번째 줄: @@
   https://arxiv.org/abs/1409.2944
-===code?===
+===code===
 저자 홈페이지: http://www.wanghao.in

@@ 5번째 줄: / 5번째 줄: @@
 ===code?===
 저자 홈페이지: http://www.wanghao.in
 공식 프로젝트 페이지: http://www.wanghao.in/CDL.htm
 저자의 다른 논문들(2016):
@@ 15번째 줄: / 16번째 줄: @@
 * Natural parameter networks: a class of probabilistic neural networks
 * Collaborative recurrent autoencoder: recommend while learning to fill in the blanks
 ===intro===