kkbox
play

KKBOX Ann Chen, York Tsai 2017/04/20 This - PowerPoint PPT Presentation

Sep 23, 2022 •146 likes •599 views

KKBOX Ann Chen, York Tsai 2017/04/20 This presentation is provided on a strictly private and confidential basis for information purposes only. Machine Learning algorithms behind Personalized

  1. KKBOX 利用機器學習讓用 戶 越聽越多元 Ann Chen, York Tsai 2017/04/20 This presentation is provided on a strictly private and confidential basis for information purposes only.

  2. Machine Learning algorithms behind Personalized Recommenders This presentation is provided on a strictly private and confidential basis for information purposes only.

  3. This presentation is provided on a strictly private and confidential basis for information purposes only.

  4. Simple Complex Collaborative Learn to Rank Persona Aware Filtering Attribute Based Context Aware Interactive Cold Start Recommender This presentation is provided on a strictly private and confidential basis for information purposes only.

  5. Click/Play Prediction ● Regression ● Classification ● Distance in Feature Space Learn to Rank Content User Understanding Understanding ● Embedding ● User Profiling ● Classification ● Embedding ● Topic Mining ● Popularity ● Trending This presentation is provided on a strictly private and confidential basis for information purposes only.

  6. This presentation is provided on a strictly private and confidential basis for information purposes only.

  7. This presentation is provided on a strictly private and confidential basis for information purposes only.

  8. This presentation is provided on a strictly private and confidential basis for information purposes only.

  9. Serendipity/Novelty Diversity Precision This presentation is provided on a strictly private and confidential basis for information purposes only.

  10. Click/Play Prediction ● Regression ● Classification ● Distance in Feature Space Learn to Rank Content User Understanding Understanding ● Embedding ● User Profiling ● Classification ● Embedding ● Topic Mining ● Popularity ● Trending This presentation is provided on a strictly private and confidential basis for information purposes only.

  11. This presentation is provided on a strictly private and confidential basis for information purposes only.

  12. Collaborative Filtering This presentation is provided on a strictly private and confidential basis for information purposes only.

  13. Word2Vec - “ The results, to our own surprise, show that the buzz is fully justified, as the context-predicting models obtain a thorough and resounding victory against their count-based counterparts. ” - Marco et al. “You should know the word by the company it keeps” (Firth J.R.) This presentation is provided on a strictly private and confidential basis for information purposes only.

  14. This presentation is provided on a strictly private and confidential basis for information purposes only.

  15. This presentation is provided on a strictly private and confidential basis for information purposes only.

  16. This presentation is provided on a strictly private and confidential basis for information purposes only.

  17. DeepWalk (Bryan Perozzi, Rami Al-Rfou& Steven Skiena, 2014 ) This presentation is provided on a strictly private and confidential basis for information purposes only.

  18. 青花瓷 給我一首歌的時 珊瑚海 我不配 間 黃金甲 霍元甲 雙截棍 天地一鬥 This presentation is provided on a strictly private and confidential basis for information purposes only.

  19. HipHop - Boom Bap, Urban, Trap Rock -Rock & Roll, Folk/Indie Pop, Punk Grunge, Dream Pop Electronic - House, Techno, Trap/Twerk Genre / Sub-genre Tempo Contex t ( 咖啡館 , 睡前 ...) Instrument Mood ( 放鬆 , 暢快 ...) Topic Vocal Activity ( 飛輪 , 旅遊 ...) Audio attributes (extract from playlist title) Sound This presentation is provided on a strictly private and confidential basis for information purposes only.

  20. Learn the relationships between lebals/latent factors and audio signals ● Multimodal ● Multiple objective This presentation is provided on a strictly private and confidential basis for information purposes only.

  21. This presentation is provided on a strictly private and confidential basis for information purposes only.

  22. Learn to rank - Regression or Classification This presentation is provided on a strictly private and confidential basis for information purposes only.

  23. Serendipity/ Novelty Personal Preferences Trending/ Popular Context Social Factors

  24. Click/Play Prediction ● Regression ● Classification ● Distance in Embedding Space Learn to Rank Content User Understanding Understanding ● Embedding ● User Profiling ● Classification ● Embedding ● Topic Mining ● Popularity ● Trending This presentation is provided on a strictly private and confidential basis for information purposes only.

  25. Challenges This presentation is provided on a strictly private and confidential basis for information purposes only.

  26. Heterogeneous Data Sources ● Database ● Log ● Streaming ● Static Dataset ● API This presentation is provided on a strictly private and confidential basis for information purposes only.

  27. Dynamic Workload This presentation is provided on a strictly private and confidential basis for information purposes only.

  28. Heterogeneous Resource Requirement ● CPU ● GPU ● Memory ● SSD This presentation is provided on a strictly private and confidential basis for information purposes only.

  29. Make Development Cycle Easy Feature Definition Deploy and Data A/B Testing Inspection Model Training and Hypothesis Verification This presentation is provided on a strictly private and confidential basis for information purposes only.

  30. Infrastructure in KKBOX Research Center This presentation is provided on a strictly private and confidential basis for information purposes only.

  31. This presentation is provided on a strictly private and confidential basis for information purposes only.

  32. Infrastructure in KKBOX RDC ● ● ● ● ● ● ● ● This presentation is provided on a strictly private and confidential basis for information purposes only.

  33. Infrastructure as Code (IaC) ● ● ● ● This presentation is provided on a strictly private and confidential basis for information purposes only.

  34. S3: Data Storage This presentation is provided on a strictly private and confidential basis for information purposes only.

  35. ETL and Data Format This presentation is provided on a strictly private and confidential basis for information purposes only.

  36. Presto: Data Analytic Engine ● ● ● ● ● This presentation is provided on a strictly private and confidential basis for information purposes only.

  37. Spark: Computing Framework ● ● ● ● ● ○ ○ ○ This presentation is provided on a strictly private and confidential basis for information purposes only.

  38. Dashboard ● ● ● ● This presentation is provided on a strictly private and confidential basis for information purposes only.

  39. Data Pipeline Examples This presentation is provided on a strictly private and confidential basis for information purposes only.

  40. Nearest Neighbors for Songs ● ● ● ● ● ● ● ● This presentation is provided on a strictly private and confidential basis for information purposes only.

  41. Predict Songs User Might Listen Again ● ● ● ● ● ● This presentation is provided on a strictly private and confidential basis for information purposes only.

  42. Genre Classification ● ● ● ● This presentation is provided on a strictly private and confidential basis for information purposes only.

  43. Spark with S3 Tips ● ● spark.hadoop.fs.s3a.connection.maximum ● ● .set("spark.hadoop.mapreduce.fileoutputcommitter.algorithm.version", "2") spark.speculation ● .set("spark.hadoop.fs.s3a.readahead.range", "512M") ● ● This presentation is provided on a strictly private and confidential basis for information purposes only.

  44. EC2 Tips ● ● ● ● ● ● ● This presentation is provided on a strictly private and confidential basis for information purposes only. ●

  45. Thank You! This presentation is provided on a strictly private and confidential basis for information purposes only.

More recommend