Upgrade to Pro — share decks privately, control downloads, hide ads and more …

オンライン広告関連の論文を50本くらい雑に紹介する AdKDD編 / adkdd-all

todesking
December 13, 2019
2k

オンライン広告関連の論文を50本くらい雑に紹介する AdKDD編 / adkdd-all

todesking

December 13, 2019
Tweet

Transcript

  1. ΦϯϥΠϯ޿ࠂؔ࿈ͷ࿦จΛ
    50ຊ͘Β͍ࡶʹ঺հ͢Δ
    AdKDDฤ
    @todesking

    View full-size slide

  2. ࣗݾ঺հ
    • @todesking

    • DSP/ΞυωοτϫʔΫͷձࣾ
    ͰػցֶशΛ΍͍ͬͯΔ

    • ೔ࠒͷٕज़ௐ͕ࠪ଍Γͳ͍ͳ
    ͱࢥͬͨͷͰAdKDDͷ࿦จΛ
    ோΊͯΈ·ͨ͠

    View full-size slide

  3. AdKDD/TargetADͱ͸
    • KDD: Knowledge Discovery and Data Mining

    • σʔλϚΠχϯά΍ػցֶशʹؔ͢ΔԠ༻తͳ࿩୊Λѻ͏ࠃࡍ
    ձٞ

    • AdKDD: ΦϯϥΠϯ޿ࠂؔ࿈ٕज़Λѻ͏෼Պձ

    • TargetAd: WSDMͱ͍͏ࠃࡍձٞʹ͓͍ͯΦϯϥΠϯ޿ࠂؔ࿈ٕज़
    Λѻ͏෼Պձ

    • ࠷ۙ͸AdKDDʹٵऩ͞Εͨ?

    • ࠾୒࿦จ਺͸ຖ೥10ຊऑ

    • KDDຊମʹ΋޿ࠂܥ࿦จ͕͋Δ(੗Έ෼͚͸Ṗ)

    View full-size slide

  4. ஫ҙ
    • ཁ໿͸΄΅AbstractோΊ͚ͨͩɺਖ਼֬͞͸อূ͠·ͤΜ

    • ؒҧ͍ͷࢦఠ͸׻ܴ

    • ۙ೥ͷ࿦จ͸ެࣜαΠτ͔ΒPDFͷ௚ϦϯΫ͕͋Δ

    • Keynote/Invited talk͸লུ

    View full-size slide

  5. AdKDD 2014
    https://dblp.org/db/conf/kdd/adkdd2014

    View full-size slide

  6. A Dynamic Pricing Model for Unifying
    Programmatic Guarantee and Real-Time
    Bidding in Display Advertising
    • publisher͕impΛചΔʹ͸RTBͱguranteed contractͱ͍
    ͏ೋͭͷํ๏͕͋Δ

    • publisherͷऩӹΛදݱ͢Δ਺ཧϞσϧΛ࡞Γɺೋͭͷํ
    ๏΁ͷ༧ࢉ഑෼Λ࠷దԽ͢Δ

    View full-size slide

  7. Multi-Touch Attribution Based Budget
    Allocation in Online Advertising
    • Ωϟϯϖʔϯશମͷ༧ࢉ͕ܾ·͓ͬͯΓɺԼҐͷΩϟϯ
    ϖʔϯʹ഑෼͍ͨ͠

    • ໨ඪ΁ͷߩݙʹԠͯ͡഑෼͍ͨ͠ˠlast-touch/multi-
    touch attributionͷ৘ใΛݩʹׂͯ͠Γ౰ͯΔํ๏Λ࣮૷
    ͨ͠

    View full-size slide

  8. Pleasing the advertising oracle:
    Probabilistic prediction from sampled,
    aggregated ground truth
    • ޿ࠂΩϟϯϖʔϯͷηάϝϯτਫ਼౓Λୈࡾऀ͕ධՁ͢Δ
    έʔε

    • ධՁͱͯ͠ɺ޿ࠂΛݟͨϢʔβͷσϞάϥଐੑͷׂ߹͕
    ϑΟʔυόοΫ͞ΕΔ

    • ݸผϢʔβͷଐੑʹ͍ͭͯͷ৘ใ͸ಘΒΕͳ͍

    • ͜ͷ৘ใΛ࢖ͬͯηάϝϯτͷਫ਼౓Λվળ͢Δ

    View full-size slide

  9. The PlaceIQ Analytic Platform: Location
    Oriented Approaches to Mobile Audiences
    • PlaceIQͱ͍͏Ґஔϕʔεͷ޿ࠂ෼ੳπʔϧͷ঺հ

    View full-size slide

  10. Practical Lessons from Predicting
    Clicks on Ads at Facebook
    • ஶ໊ͳΦϯϥΠϯ޿ࠂ࿦จ

    • FBͷCTR༧ଌϞσϧͷ࣮૷ʹ͍ͭͯ

    View full-size slide

  11. iPinYou Global RTB Bidding
    Algorithm Competition Dataset
    • iPinYouͱ͍͏DSP͕ओ࠵ͨ͠RTBΞϧΰϦζϜͷίϯς
    ετͰ࢖༻͞Εͨσʔληοτʹ͍ͭͯ

    • ͜ͷσʔληοτ͸RTBܥϩδοΫͷ࣮ݧʹΑ͘࢖ΘΕͯ
    ͍Δ

    View full-size slide

  12. Can Television Advertising Impact Be Measured
    on the Web? Web Spike Response as a Possible
    Conversion Tracking System for Television.
    • ςϨϏࢹௌऀ͸ಉ࣌ʹΠϯλʔωοτ΋΍ΔͷͰɺςϨϏ
    CMͷࢹௌ͕े਺ඵޙͷωοτ׆ಈʹӨڹΛ༩͑Δ

    • ͜ͷΑ͏ͳӨڹΛܭଌ͢ΔͨΊͷwebϕʔεγεςϜΛ։
    ൃͨ͠

    View full-size slide

  13. TargetAd 2016
    https://sites.google.com/site/targetad2016/program

    View full-size slide

  14. TargetAd 2016
    https://sites.google.com/site/targetad2016/program

    View full-size slide

  15. TargetAd 2016
    https://sites.google.com/site/targetad2016/program

    View full-size slide

  16. Distributed Representations of Web
    Browsing Sequences for Ad Targeting
    • Ϣʔβͷϒϥ΢δϯάཤྺΛಛ௃ྔԽ͍ͨ͠

    • ϢʔβΛparagraphɺURLΛwordʹݟཱͯͯParagraph
    Vector(W2Vͷ೿ੜɺύϥάϥϑͷembeddingΛಘΒΕΔ)
    Λߏங͢Δͱ͍͏ͷΛલճ΍ͬͨ

    • Backward PV-DMͱ͍͏վྑ൛ͷఏҊ

    View full-size slide

  17. Sponsored Video Advertising: Filling the
    Void between Online and TV Advertising
    • Ϣʔβ͕ಈըΛࢹௌ͍ͯ͠Δͱ͖ʹ޿ࠂΛݟ͍ͤͨɺݟͤ
    ·͘Γ͍ͨͱ͍͏ڧ͍ؾ͕࣋ͪ͋Δ

    • ಈըʹؔ࿈͢Δ޿ࠂͷ՝ۚମܥʹ͍ͭͯͷఏҊ

    View full-size slide

  18. On Spatial-Aware Viral Marketing
    for Location-based Advertisements,
    • ৽͍͠ళΛ։͖͍ͨ৔߹ɺళͷ৔ॴɺޱίϛͷޮՌɺ෺
    ཧ޿ࠂͷ৔ॴͳͲͷཁૉ͕ച্͛ʹӨڹΛ༩͑Δɻ

    • ͦͷΑ͏ͳ੍໿ͷԼͰ࠷దͳީิΛબͿํ๏ΛఏҊ

    View full-size slide

  19. Keynote: Come back soon!
    Estimating return times for users
    • ߦಈཤྺΛݩʹɺϢʔβ͕࣍ʹ͜ͷαʔϏεΛ࢖͏ͷ͸
    ͍͔ͭਪఆ͍ͨ͠

    • ੜଘ෼ੳɺ఺աఔɺLSTM

    View full-size slide

  20. Building a Bi-directed Recommendation
    System for Mobile Users and App-install
    Ad Campaigns
    • ϞόΠϧΞϓϦʹؔ͢Δ૒ํ޲Ϩίϝϯσʔγϣϯγες
    Ϝ

    • ΞϓϦͷ޿ࠂΩϟϯϖʔϯʹରͯ͠ɺޮՌతͳϢʔβηά
    ϝϯτΛਪન

    • Ϣʔβʹରͯ͠ɺڵຯΛ࣋ͪͦ͏ͳΞϓϦΛਪન

    View full-size slide

  21. Finding Needle in a Million Metrics:
    Anomaly Detection in a Large-scale
    Computational Advertising Platform
    • ޿ࠂϓϥοτϑΥʔϜʹ͓͚Δҟৗݕ஌

    • Ξυϑϥ΢υͷจ຺Ͱ͸ͳ͘ɺγεςϜͷҟৗΛݕग़͢Δ
    ͷ͕໨త

    • ҆ఆͨ͠ύϑΥʔϚϯεΛग़͍ͯ͠ΔΩϟϯϖʔϯΛ؍ଌ

    • ύϑΥʔϚϯεʹҟৗ͕͋ͬͨΒɺγεςϜଆʹݪҼ͕
    ͋ΔՄೳੑ͕ߴ͍

    View full-size slide

  22. Preserving Privacy in Geo-
    Targeted Advertising
    • ஍ҬͱऩೖΛ࢖ͬͨλʔήςΟϯά޿ࠂΛߟ͑Δ

    • ߈ܸऀ͸τϐοΫϞσϦϯά΍ػցֶशΛ࢖͏͜ͱͰɺ͋
    Δ޿ࠂ͕දࣔ͞ΕͨϢʔβͷ஍Ҭ΍ऩೖΛਪఆͰ͖Δ

    • ഑৴͞ΕΔ޿ࠂʹϊΠζΛࠞͥΔ͜ͱͰ͜ͷΑ͏ͳਪఆΛ
    ๦͛Δ

    View full-size slide

  23. Invited talk: User
    Recommendation in Instagram
    • Instagramʹ͓͚Δ͓͢͢ΊϢʔβͷ࢓૊Έ

    • ιʔγϟϧάϥϑͷߏ଄΍ڵຯͷ͋ΔτϐοΫΛ࢖ͬͨਪ

    View full-size slide

  24. Demographic Prediction of Web
    Requests from Labeled Aggregate Data
    • Ϣʔβू߹ʹରͯ͠ɺσϞάϥଐੑͷׂ߹Λ஌Δ͜ͱ͕Ͱ
    ͖Δͱ͢Δ

    • ͜ͷ৘ใΛ࢖ͬͯϢʔβ୯ҐͰͷଐੑਪఆΛ͢Δ

    • Williams+, "Pleasing the advertising oracle: Probabilistic
    prediction from sampled, aggregated ground truth",
    AdKDD2014 ͷൃల

    View full-size slide

  25. Toward Personalized Product Search for
    eCommerce Sites: A Case Study in Yahoo!
    Taiwan
    • Y!୆࿷ͷeίϚʔεαΠτʹ͓͚Δݕࡧ

    • ଟ͘ͷݕࡧΫΤϦ͸ҰൠతͩͬͨΓᐆດͰɺϢʔβ͸๲େ
    ͳ݁Ռͷத͔Β໨తͷ΋ͷΛ୳͞Ͷ͹ͳΒͳ͍

    • ςΩετ৘ใ͚ͩͰͷϚονϯάͰ͸ෆे෼ͳͷͰɺϢʔ
    βͷڵຯʹԠͯ͡ύʔιφϥΠζͨ͠ϥϯΩϯάΛߦ͏

    View full-size slide

  26. AdKDD & TargetAd
    2017

    View full-size slide

  27. AdKDD & TargetAd 2017
    https://adkdd-targetad.wixsite.com/adkddtargetad2017/accepted-papers

    View full-size slide

  28. AdKDD & TargetAd 2017
    https://adkdd-targetad.wixsite.com/adkddtargetad2017/accepted-papers

    View full-size slide

  29. AdKDD & TargetAd 2017
    https://adkdd-targetad.wixsite.com/adkddtargetad2017/accepted-papers

    View full-size slide

  30. Attribution Modeling Increases Efficiency of Bidding in
    Display Advertising
    • Criteo

    • ΞτϦϏϡʔγϣϯϞσϧΛ࢖ͬͨbidઓུ

    • 2nd price auctionʹ͓͚Δ࠷దઓུ͸ͦͷϦΫΤετͷՁ
    ஋Λೖࡳֹͱ͢Δ͜ͱ͕ͩɺਖ਼֬ͳՁ஋ΛٻΊΔͨΊʹ
    ͸ΞτϦϏϡʔγϣϯϞσϧΛߟྀ͢Δඞཁ͕͋Δ

    View full-size slide

  31. Blacklisting the Blacklist in
    Online Advertising
    • Dstillery

    • RTBʹ͓͍ͯ͸publisherଆ͕ϒϥοΫϦετΛ͍࣋ͬͯͯ
    ಛఆͷ޿ࠂΛ͸͘͜͡ͱ͕͋Δ

    • ϒϥοΫϦετͷ಺༰͸ඇެ։

    • ແବͳbidΛආ͚ΔͨΊϒϥοΫϦετͷ಺༰Λਪఆ͢Δ
    ػೳΛ࡞ͬͨ

    • γεςϜෛՙ͕ݮΓɺউ཰͕޲্ͨ͠

    View full-size slide

  32. Anti-Ad Blocking Strategy:
    Measuring its True Impact
    • Adobe Research

    • AdBlockΛݕग़ͯ͠ϢʔβʹϝοηʔδΛग़͢ઓུ͕͋Δ
    (՝ۚίʔε΁ͷ༠ಋͳͲ)

    • ͜ͷઓུ͸Ͳͷఔ౓༗ޮͳͷ͔ௐ΂ΔͨΊͷ౷ܭతख๏
    ΛఏҊ

    • ͳΜΒ͔ͷࣄ৘ʹΑΓɺී௨ʹA/Bςετ͢ΔΘ͚ʹ͸͍
    ͔ͳ͍Β͍͠

    View full-size slide

  33. MM2RTB: Bringing Multimedia
    Metrics to Real-Time Bidding
    • ޿ࠂͷ࣭ΛධՁ͢ΔͨΊͷmultimedia metricsͱ͍͏ͷ͕
    ͋Δ

    • contextual relevance, visual saliency, ad memorability

    • ϖʔδ಺ʹෳ਺ͷ޿ࠂεϩοτ͕͋Δͱ͖ɺmultimedia
    metricsΛߟྀͯ͠޿ࠂͷ૊Έ߹ΘͤΛબ୒͢Δख๏ͷఏ
    Ҋ

    View full-size slide

  34. Data-Driven Reserve Prices for Social
    Advertising Auctions at LinkedIn
    • ηΧϯυϓϥΠεΦʔΫγϣϯʹ͓͚Δ࠷దͳϦβʔϒϓ
    ϥΠεΛٻΊΔํ๏Λ2छྨ(Ϣʔβ/ηάϝϯτϨϕϧ)ఏ
    Ҋ

    • ࣮γεςϜʹ࣮૷ͯ͠ޮՌΛݕূͨ͠

    View full-size slide

  35. Optimal Reserve Price for Online Ads
    Trading Based on Inventory Identification
    • Alibaba+Yahoo

    • ͜Ε΋reserve priceͷܾఆઓུʹؔ͢Δ࿩

    View full-size slide

  36. Ranking and Calibrating Click-Attributed
    Purchases in Performance Display
    Advertising
    • click-attributionͰCV՝ۚϞσϧͷ޿ࠂ഑৴

    • bidͷࡍʹ͸ҎԼͷखॱ͕ඞཁ

    • CV͞Ε΍͍͢޿ࠂͷީิΛྻڍ(ΞτϦϏϡʔγϣϯͷ
    ੍໿ͰɺΫϦοΫ཰΋ߴ͍ඞཁ͕͋Δ)

    • CV཰ʹԠͯ͡ೖࡳֹۚΛܾఆ(ਖ਼֬ͳCV཰ͷਪఆ͸ࠔ
    ೉)

    • ϥϯΩϯά+ΩϟϦϒϨʔγϣϯʹΑͬͯ͜ͷ໰୊Λղ͘

    View full-size slide

  37. Cost-sensitive Learning for Utility
    Optimization in Online Advertising
    Auctions
    • CTR/CVR༧ଌʹ͸ɺhighly non-uniform misprediction
    cost͕ଘࡏ͢Δ

    • UtilityͷΑ͏ͳࢦඪΛ࢖ͬͯੑೳධՁ͢Δख๏͕ఏҊ͞Ε
    ͍ͯΔ

    • ֶश࣌ʹ΋UtilityΛߟྀͯ͠࠷దԽ͍ͨ͠

    • log lossʹॏΈΛ͚ͭΔ͜ͱͰੑೳվળ͢Δख๏ͷఏҊ

    View full-size slide

  38. A Practical Framework of Conversion Rate
    Prediction for Online Display Advertising
    • CVR༧ଌʹؔ͢Δ༷ʑͳख๏ͷఏҊ

    • Over prediction΁ͷରԠ

    • delayed feedbackΛߟֶྀͨ͠श

    • ΞτϦϏϡʔγϣϯϞσϧΛߟྀͨ͠ೖࡳֹิਖ਼

    View full-size slide

  39. An Ensemble-based Approach to Click-
    Through Rate Prediction for Promoted
    Listings at Etsy
    • Etsyͷlisting adͰ࢖ΘΕ͍ͯΔCTRਪఆγεςϜͷ঺հ

    View full-size slide

  40. Profit Maximization for Online
    Advertising Demand-Side Platform
    • CPC/CPAϞσϧʹ͓͍ͯɺDSPͷརӹΛ࠷େԽ͍ͨ͠

    • ࠷దͳೖࡳઓུ(Ͳͷ޿ࠂΛ͍͘ΒͰೖࡳ͢Δ͔)ʹ͍ͭͯ
    ߟ࡯

    • Ωϟϯϖʔϯ༧ࢉͱϖʔγϯάɺλʔήςΟϯάɺimpڙ
    څྔͷ੍໿͕͋Δ

    • ࣮༻తʹղ͚ΔϞσϧΛఏҊ

    View full-size slide

  41. Deep & Cross Network for
    Ad Click Predictions
    • Google

    • Deep & Cross networkͱ͍͏DLͷϞσϧΛఏҊ

    • ަޓ࡞༻Λཅʹѻ͑Δ

    • CTRਪఆʹར༻ͯ͠ߴੑೳͩͬͨ

    View full-size slide

  42. AdKDD & TargetAd 2018

    View full-size slide

  43. AdKDD & TargetAd 2018
    https://adkdd-targetad.wixsite.com/2018/accepted-papers

    View full-size slide

  44. AdKDD & TargetAd 2018
    https://adkdd-targetad.wixsite.com/2018/accepted-papers

    View full-size slide

  45. A Large Scale Benchmark
    for Uplift Modeling
    • Criteo

    • Uplift modelingͷݕূʹ࢖͑Δେن໛σʔλΛެ։ͨ͠

    • https://ailab.criteo.com/criteo-uplift-prediction-dataset/

    View full-size slide

  46. Deep Density Networks and
    Uncertainty in Recommender Systems
    • Taboola(޿ࠂϓϥοτϑΥʔϜΛ΍ͬͯΔձࣾ)

    • Deep Density Networksͱ͍͏ϞσϧͷఏҊ

    • content based+ڠௐϑΟϧλϦϯάͷϋΠϒϦου

    • ༧ଌͷෆ࣮֬ੑΛදݱͰ͖Δ

    • ε-greedyͱUCBΛ૊Έ߹ΘͤͨΑ͏ͳϩδοΫͰ
    exploration/exploitation͢Δ

    View full-size slide

  47. Deep Neural Net with Attention for
    Multi-channel Multi-touch Attribution
    • DNNΛ࢖ͬͨMulti-touch attributionϞσϧͷఏҊ

    View full-size slide

  48. Deep Policy Optimization for E-commerce
    Sponsored Search Ranking Strategy
    • 2018͔ΒDNNΛ࢖ͬͨൃද͕໨ཱͪ·͢Ͷ

    • εϙϯαʔυαʔνͷϥϯΩϯά໰୊Λਂ૚ڧԽֶशͰ
    ղ͘

    View full-size slide

  49. Designing Experiments to Measure
    Incrementality on Facebook
    • Facebookʹ͓͚Δincrementalityܭଌख๏ͷղઆ

    • upliftͱincrementalityͷҧ͍͕Ṗ

    • ୯ͳΔA/BςετΑΓॊೈͳ࣮ݧ͕Մೳ

    View full-size slide

  50. Dynamic Hierarchical Empirical Bayes: A
    Predictive Model Applied to Online
    Advertising
    • Adobe

    • εϙϯαʔυαʔνʹ͓͚ΔCTR/CVRਪఆϞσϧ

    • ֊૚ϕΠζʹ͓͍ͯɺσʔλʹԠͯ͡ಈతʹ֊૚ߏ଄Λܾ
    ఆ͢ΔΑ͏ͳϞσϧͷఏҊ

    View full-size slide

  51. Forecasting Granular Audience
    Size for Online Advertising
    • ಛఆͷηάϝϯτʹ޿ࠂ഑৴͢Δͱ͖ɺࣄલʹΦʔσΟΤ
    ϯεͷαΠζ͕஌Γ͍ͨ

    • Frequent Itemset Miningͱ͍͏ख๏ΛݩʹɺͦͷΑ͏ͳਪ
    ఆΛ͢Δख๏ΛఏҊ

    View full-size slide

  52. Mini-Batch AUC
    Optimization
    • ֶश࣌ʹAUCΛ௚઀࠷దԽ͍͕ͨ͠ɺେن໛σʔληο
    τʹ͓͍ͯ͸೉͔ͬͨ͠

    • ϛχόονΛ࢖͏͜ͱͰߴ଎ͳֶशΛՄೳͱ͢Δख๏Λ
    ఏҊ

    View full-size slide

  53. Optimal Bidding, Allocation and Budget
    Spending for a Demand Side Platform
    Under Many Auction Types
    • ৽͍͠ೖࡳઓུϩδοΫͷఏҊ

    • 1st/2nd price auctionʹରԠɺ༧ࢉ੍໿΍ΫϥΠΞϯτ͝
    ͱͷر๬ΛߟྀͰ͖Δ

    View full-size slide

  54. AdKDD 2019
    https://www.adkdd.org/2019-papers-and-talks

    View full-size slide

  55. AdKDD 2019
    https://www.adkdd.org/2019-papers-and-talks

    View full-size slide

  56. AdKDD 2019
    https://www.adkdd.org/2019-papers-and-talks

    View full-size slide

  57. Time-Aware Prospective Modeling of
    Users for Online Display Advertising
    • Yahoo

    • ࣌ܥྻΛߟྀͨ͠ϢʔβߦಈϞσϦϯάʹΑΓɺϢʔβͷ
    ҙਤΛߟྀͨ͠ਪఆΛߦ͏DNNϞσϧͷఏҊ

    View full-size slide

  58. Graphing Crumbling
    Cookies
    • ΫϩεσόΠετϥοΩϯάΛ͍͕ͨ͠ɺۙ೥cookieͷ࢖
    ༻੍ݶ͕ݫ͘͠ͳ͍ͬͯΔ(ITP౳)

    • Browser fingerprintingͳͲͷΘΔ͍ٕज़Λ࢖Θͣ͜ͷঢ়گ
    ʹରԠ͢ΔͨΊͷख๏ΛఏҊ

    View full-size slide

  59. Causally Driven Incremental Multi Touch
    Attribution Using a Recurrent Neural
    Network
    • JD.com(ژ౦঎৓): தࠃͷڊେECاۀ

    • multi-touch attributionΛRNNͰϞσϦϯά͢Δख๏Λఏ
    Ҋ

    • ͜ͷϞσϧ͸࣮ࡍ࢖ΘΕ͍ͯΔΒ͍͠

    View full-size slide

  60. Combinatorial Keyword
    Recommendations for Sponsored Search
    with Deep Reinforcement Learning
    • Baidu

    • ਂ૚ڧԽֶशͰϨίϝϯμ࡞Δ΍ͭ

    View full-size slide

  61. Feasible Bidding Strategies
    through Pure Exploration Bandits
    • ೖࡳઓུͷީิ͕େྔʹ͋Γɺ͍͍΍ͭΛબͼ͍ͨ

    • ύϥϝλͷ૊Έ߹ΘͤͰ͔ͳΓީิ͕ଟ͍(~100)ͨΊɺී
    ௨ʹόϯσΟοτΞϧΰϦζϜΛ࢖͏ͷ͸ඇޮ཰

    • ʮͦͦ͜͜Αͦ͞͏ʯͳީิΛߴ଎ʹൃݟ͢ΔͨΊͷख๏
    ΛఏҊ

    View full-size slide

  62. In-app Purchase Prediction Using
    Bayesian Personalized DwellDay Ranking
    • ژେ+CyberAgent

    • ΞϓϦ಺՝ۚͦ͠͏ͳϢʔβΛਪఆ͢ΔλεΫ

    • Bayesian personalized rankingΛݩʹͨ͠ख๏ΛఏҊ

    View full-size slide

  63. Learning from Multi-User Activity
    Trails for B2B Ad Targeting
    • Yahoo

    • B2Bʹ͓͚Δ঎඼ͷߪೖϓϩηε͸ɺෳ਺ͷਓ͕ؒؔΘͬ
    ͨෳࡶͳ෺ʹͳΔ

    • ͜ͷΑ͏ͳϓϩηεʹରԠͨ͠޿ࠂͷ഑৴ઓུ͕ඞཁ

    • ಉ͡૊৫ʹ͍ΔΦʔσΟΤϯεΛਪఆɺͦΕΒͷߦಈΛ૯
    ߹ͯ͠CV༧ଌ͢ΔϞσϧΛఏҊ

    View full-size slide

  64. Modeling Advertiser Bidding Behaviors in
    Google Sponsored Search with a Mirror
    Attention Mechanism
    • SSPଆ͔ΒೖࡳऀͷߦಈΛϞσϦϯά͢Δ

    • ৽ػೳΛϦϦʔεͨ͠ࡍɺͦΕ͕௕ظతʹ޿ࠂओʹͲͷΑ
    ͏ͳӨڹΛٴ΅͔͢஌Γ͍ͨ…… ͱ͍͏ͷ͕Ϟνϕʔ
    γϣϯΒ͍͠

    View full-size slide

  65. Optimal bidding: a dual
    approach
    • ৽͍͠ೖࡳઓུͷఏҊ

    View full-size slide

  66. ·ͱΊ
    • ޿ࠂܥ࿦จ͕ͨ͘͞Μ͋Δ͜ͱ͕Θ͔Γ·ͨ͠

    • ͍͔͕Ͱ͔ͨ͠?

    • དྷ೥಄͘Β͍ʹ޿ࠂܥ࿦จಡॻձΛ΍Γ͍ͨͷͰɺڵຯ͋
    Δਓ͸͓੠͕͚͍ͩ͘͞

    View full-size slide