此篇論文中，我們提出了一種在線上音樂播放中使用的動態權重機
制。基於隱性因子模型之假設，使用者、歌曲、歌手皆是由一組隱性
空間中的向量表示。給定一使用者以及他的近期播放紀錄，即可判斷
他目前的興趣是傾向近期偏好或是長期偏好。同其他隱性因子模型，
此機制可在不使用內容資料的情形下訓練。這在使用網路播放紀錄當
做資料庫時是一個利基，因內容資料相對較難取得。在數個 last.fm 資
料集上的實驗結果顯示此方法確實有效。
關鍵字：音樂推薦，動態興趣，隱性因子向量模型，機器學習，梯度
上升

In this paper, we propose a dynamic weight tuning scheme for online mu-
sic recommendation. Based on a latent factor model, songs, artists, and users
are mapped into a latent space. Then, given each user’s recent songs we can
determine his current interest for music, which either similar to his past be-
havior or more like recent ones. Like latent factor based models, this scheme
can be trained without content information, which is a benefit when adopting
internet radios as data source. Experimental results on the last.fm collections
show that our proposed method is effective.
Keywords: Music Recommendation, Dynamic Interests, Latent Factor Model,
Machine Learning, Gradient Ascent

誌謝 i
摘要 ii
Abstract iii
1 Introduction 1
2 Related Work 3
3 Proposed method 5
3.1 Latent Factor Vector Model . . . . . . . . . . . . . . . . . . . . . . . . . 5
3.2 Learning Latent Factor Vectors . . . . . . . . . . . . . . . . . . . . . . . 6
3.3 Dynamic Weighting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
3.4 Features . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
3.5 Heuristic Filtering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
3.6 Learning Weight Vectors . . . . . . . . . . . . . . . . . . . . . . . . . . 11
3.7 Postprocessing with SVM . . . . . . . . . . . . . . . . . . . . . . . . . . 11
4 Experiments 12
4.1 Experiment Settings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
4.2 Evaluation Metric . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
4.3 Datasets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
4.4 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
4.5 Feature Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
5 Conclusions and Future Work 18
5.1 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
5.2 Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
Bibliography 19

[1] N. Aizenberg, Y. Koren, and O. Somekh. Build your own music recommender by
modeling internet radio streams. In Proceedings of the 21st international conference
on World Wide Web, pages 1–10. ACM, 2012.
[2] M. A. Bartsch and G. H. Wakefield. Audio thumbnailing of popular music using
chroma-based representations. Multimedia, IEEE Transactions on, 7(1):96–104,
2005.
[3] Y. Bengio, J.-S. Sen&;#233;cal, et al. Quick training of probabilistic neural nets by impor-
tance sampling. In AISTATS Conference, 2003.
[4] D. Bogdanov, M. Haro, F. Fuhrmann, A. Xamb&;#243;, E. G&;#243;mez, and P. Herrera. Se-
mantic audio content-based music recommendation and visualization based on user
preference examples. Information Processing &; Management, 49(1):13–33, 2013.
[5] J. Bu, S. Tan, C. Chen, C. Wang, H. Wu, L. Zhang, and X. He. Music recommen-
dation by unified hypergraph: combining social media information and music con-
tent. In Proceedings of the international conference on Multimedia, pages 391–400.
ACM, 2010.
[6] O. Celma. Music Recommendation and Discovery in the Long Tail. Springer, 2010.
19[7] C.-C. Chang and C.-J. Lin. LIBSVM: A library for support vector machines. ACM
Transactions on Intelligent Systems and Technology, 2:27:1–27:27, 2011. Software
available at http://www.csie.ntu.edu.tw/~cjlin/libsvm.
[8] P. Grosche, M. M&;#252;ller, and J. Serr&;#224;. Audio content-based music retrieval. Dagstuhl
Follow-Ups, 3, 2012.
[9] N. Hariri, B. Mobasher, and R. Burke. Using social tags to infer context in hybrid
music recommendation. In Proceedings of the twelfth international workshop on
Web information and data management, pages 41–48. ACM, 2012.
[10] M. Kaminskas, F. Ricci, and M. Schedl. Location-aware music recommendation
using auto-tagging and hybrid matching. In Proceedings of the 7th ACM conference
on Recommender systems, pages 17–24. ACM, 2013.
[11] Y. Koren, R. Bell, and C. Volinsky. Matrix factorization techniques for recom-
mender systems. Computer, 42(8):30–37, 2009.
[12] last.fm. http://last.fm.
[13] Y.-I. Song, J.-T. Lee, and H.-C. Rim. Word or phrase?: learning which unit to stress
for information retrieval. In Proceedings of the Joint Conference of the 47th Annual
Meeting of the ACL and the 4th International Joint Conference on Natural Language
Processing of the AFNLP: Volume 2-Volume 2, pages 1048–1056. Association for
Computational Linguistics, 2009.
[14] X. Wang, D. Rosenblum, and Y. Wang. Context-aware mobile music recommenda-
tion for daily activities. In Proceedings of the 20th ACM international conference
on Multimedia, pages 99–108. ACM, 2012.
20[15] X. Wu, Q. Liu, E. Chen, L. He, J. Lv, C. Cao, and G. Hu. Personalized next-song
recommendation in online karaokes. In Proceedings of the 7th ACM conference on
Recommender systems, pages 137–140. ACM, 2013.
[16] S. Yoshizaki, Y. Yoshitomi, C. Koro, and T. Asada. Music recommendation hybrid
system for improving recognition ability using collaborative filtering and impression
words. Artificial Life and Robotics, 18(1-2):109–116, 2013.