In this section, we investigate how the dimension of different fields affect the perfor-mance. We set user dimension range from 0 to 100, item dimension range from 0 to 100, then conduct two machine applications (multi-label classification and link prediction) to evaluate the performance. Figure 5.4 shows the performance of two machine applications
0 10 20 30 40 50 60 70 80 90 100 Figure 5.4: The performance w.r.t User-Movie dimensions on ML-1m.
w.r.t different User-Movie dimension pair on Movielens 1m dataset. From two figures, we find that the dimension of movies as 20∼ 30 can obtain the best result, the dimension higher than this range might overfit the training data. On the other hand, the dimension need of user is range 40 ∼ 70 in multi-label classification task and 35 ∼ 80 in link pre-diction task. The most interesting is that we only use the learned movie representations to train a classifier in multi-label classification task, the model without user information get the worst performance. In link prediction task which needs both user-representations and movie-representations, the model without any information get the worst performance.
Chapter 6 Conclusion
In this paper, we propose a field-aware network embedding that models the rich it-eration information from heterogeneous network. The proposed model use the “Multi-Representation Single Context” structure which models the different types of interactions separately and generates low dimension vertices representations for each field. Our model not only can learns iteration from different type without pre-process network but also deal-ing with the incomparable problem of different type of nodes. Also, the structure of our model makes us able to set different field dimension which provides more flexibility and explanation of data.
The experimental results on various networks show that it is able to improve the rep-resentation capability of machine learning applications such as network reconstruction, multi-label classification and link prediction. Moreover, our model have much improve the performance of cross-field problem such as recommendation than other network em-bedding methods.
Bibliography
[1] A. Ahmed, L. Hong, and A. J. Smola. Hierarchical geographical modeling of user locations from social media posts. In Proceedings of the 22Nd International Con-ference on World Wide Web, WWW ’13, pages 25–36, New York, NY, USA, 2013.
ACM.
[2] A. Ahmed, N. Shervashidze, S. Narayanamurthy, V. Josifovski, and A. J. Smola.
Distributed large-scale natural graph factorization. In Proceedings of the 22Nd In-ternational Conference on World Wide Web, WWW ’13, pages 37–48, New York, NY, USA, 2013. ACM.
[3] M. Belkin and P. Niyogi. Laplacian eigenmaps for dimensionality reduction and data representation. Neural Comput., 15(6):1373–1396, June 2003.
[4] S. Cao, W. Lu, and Q. Xu. Grarep: Learning graph representations with global structural information. In Proceedings of the 24th ACM International on Conference on Information and Knowledge Management, CIKM ’15, pages 891–900, New York, NY, USA, 2015. ACM.
[5] S. Chang, W. Han, J. Tang, G.-J. Qi, C. C. Aggarwal, and T. S. Huang. Hetero-geneous network embedding via deep architectures. In Proceedings of the 21th ACM SIGKDD International Conference on Knowledge Discovery and Data Min-ing, KDD ’15, pages 119–128, New York, NY, USA, 2015. ACM.
[6] C.-M. Chen, M.-F. Tsai, Y.-C. Lin, and Y.-H. Yang. Query-based music recommen-dations via preference embedding. In Proceedings of the 10th ACM Conference
on Recommender Systems, RecSys ’16, pages 79–82, New York, NY, USA, 2016.
ACM.
[7] R.-E. Fan, K.-W. Chang, C.-J. Hsieh, X.-R. Wang, and C.-J. Lin. LIBLINEAR:
A library for large linear classification. Journal of Machine Learning Research, 9:1871–1874, 2008.
[8] A. Grover and J. Leskovec. Node2vec: Scalable feature learning for networks. In Proceedings of the 22Nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, KDD ’16, pages 855–864, New York, NY, USA, 2016.
ACM.
[9] C. Guo and X. Liu. Automatic feature generation on heterogeneous graph for music recommendation. In Proceedings of the 38th International ACM SIGIR Conference on Research and Development in Information Retrieval, SIGIR ’15, pages 807–810, New York, NY, USA, 2015. ACM.
[10] F. M. Harper and J. A. Konstan. The movielens datasets: History and context. ACM Trans. Interact. Intell. Syst., 5(4):19:1–19:19, Dec. 2015.
[11] Y. Jacob, L. Denoyer, and P. Gallinari. Learning latent representations of nodes for classifying in heterogeneous social networks. In Proceedings of the 7th ACM International Conference on Web Search and Data Mining, WSDM ’14, pages 373–
382, New York, NY, USA, 2014. ACM.
[12] Y. Juan, Y. Zhuang, W.-S. Chin, and C.-J. Lin. Field-aware factorization machines for ctr prediction. In Proceedings of the 10th ACM Conference on Recommender Systems, RecSys ’16, pages 43–50, New York, NY, USA, 2016. ACM.
[13] X. Liu, Y. Yu, C. Guo, and Y. Sun. Meta-path-based ranking with pseudo relevance feedback on heterogeneous graph for citation recommendation. In Proceedings of the 23rd ACM International Conference on Conference on Information and Knowledge Management, CIKM ’14, pages 121–130, New York, NY, USA, 2014. ACM.
[14] C. Meng, R. Cheng, S. Maniu, P. Senellart, and W. Zhang. Discovering meta-paths in large heterogeneous information networks. In Proceedings of the 24th International Conference on World Wide Web, WWW ’15, pages 754–764, Republic and Canton of Geneva, Switzerland, 2015. International World Wide Web Conferences Steering Committee.
[15] B. Perozzi, R. Al-Rfou, and S. Skiena. Deepwalk: Online learning of social repre-sentations. In Proceedings of the 20th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, KDD ’14, pages 701–710, New York, NY, USA, 2014. ACM.
[16] S. Rendle and L. Schmidt-Thieme. Pairwise interaction tensor factorization for per-sonalized tag recommendation. In Proceedings of the Third ACM International Con-ference on Web Search and Data Mining, WSDM ’10, pages 81–90, New York, NY, USA, 2010. ACM.
[17] Y. Sun, J. Han, C. C. Aggarwal, and N. V. Chawla. When will it happen?: Relation-ship prediction in heterogeneous information networks. In Proceedings of the Fifth ACM International Conference on Web Search and Data Mining, WSDM ’12, pages 663–672, New York, NY, USA, 2012. ACM.
[18] J. Tang, T. Lou, J. Kleinberg, and S. Wu. Transfer learning to infer social ties across heterogeneous networks. ACM Trans. Inf. Syst., 34(2):7:1–7:43, Apr. 2016.
[19] J. Tang, M. Qu, and Q. Mei. Pte: Predictive text embedding through large-scale het-erogeneous text networks. In Proceedings of the 21th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, KDD ’15, pages 1165–1174, New York, NY, USA, 2015. ACM.
[20] J. Tang, M. Qu, M. Wang, M. Zhang, J. Yan, and Q. Mei. Line: Large-scale infor-mation network embedding. In Proceedings of the 24th International Conference
on World Wide Web, WWW ’15, pages 1067–1077, Republic and Canton of Geneva, Switzerland, 2015. International World Wide Web Conferences Steering Committee.
[21] D. Wang, P. Cui, and W. Zhu. Structural deep network embedding. In Proceedings of the 22Nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, KDD ’16, pages 1225–1234, New York, NY, USA, 2016. ACM.
[22] J. Yang and J. Leskovec. Defining and evaluating network communities based on ground-truth. In Proceedings of the ACM SIGKDD Workshop on Mining Data Se-mantics, MDS ’12, pages 3:1–3:8, New York, NY, USA, 2012. ACM.
[23] X. Yu, X. Ren, Y. Sun, Q. Gu, B. Sturt, U. Khandelwal, B. Norick, and J. Han. Per-sonalized entity recommendation: A heterogeneous information network approach.
In Proceedings of the 7th ACM International Conference on Web Search and Data Mining, WSDM ’14, pages 283–292, New York, NY, USA, 2014. ACM.