TrendOwl

登壇詳細→https://youtu.be/PB4Wg-y1JrI?t=7942

Transcript

2. .FNCFST Α͏͔Μ ͿΜͿΜ ͍ʔͪΌΜ 5ZMFS Ωϯδϣ΢ ͱΈͬ͘͢ ࣎լ ژ౎ ౦ژ ෱ౡ aa
   ࢲ͕͸ͳ͍ͯ͠·͢
   
4. IUUQTRJJUBDPNBCDTIPUBSPJUFNTGDCBDF
5. ເΛඳ͖͍ͨ
   ʮϑΝΫτνΣοΫͷຽओԽʯ
   Λ໨ࢦ͍ͯ͠·͢ʜ
6. ੍࡞෺ 🤔ϑΣΠΫχϡʔεͱ͸
7. ໋΍੓࣏ʹؔΘΔ͜ͱ΋ ۽ຊ஍਒ ೥ถબڍ
8. ϫΫνϯ͸ࣗ෼ͷҙࢥͰଧͯΔ ͷʹɺଧͭͳͱڧ੍͞Εͯ͠ ·͍ͬͯΔʜ ࣮ࡍͷମݧஊ ਌͕৘ใݯΛͭʹཔΓ͗ͯ͢ɺ ภͬͨ৘ใͰ࿩͢Α͏ʹͳΓɺ
   ձ࿩͕੒Γཱͨͳ͍ ϝϯόʔͷ5͘Μ ༑ୡͷ/͞Μ
9. ૯຿লͷൃදͰ͸ϲ݄ʹҎ্ͷਓ͸ϑΣΠΫχϡʔεʹ৮Ε͍ͯΔ
   Ҿ༻ɿIUUQTXXXTPVNVHPKQNBJO@TPTJLJKPIP@UTVTJOE@TZPIJJIPZVHBJ@IUNM 
 ʢຊݚڀձʢୈ
   ճʣࢿྉ̍ʮʰϑΣΠΫχϡʔεʱʹؔ͢ΔΞϯέʔτௐࠪ݁Ռʯʢ/3*ʣʣ
10. ίϩφʹؔ͢Δ৘ใͷਅِΛ൑அͰ͖ͳ͔ͬͨਓ͕ଟ͍܏޲ ɾίϩφՒͷ೔ຊͰ͸ɺ 
 ʮ͓౬Ͱ৽ܕίϩφ΢Πϧε͕ࢮ໓͢Δʯ 
 ΍ʮ৽ܕίϩφ΢ΠϧεͷӨڹͰ 
 τΠϨοτϖʔύʔ͕ෆ଍͢Δʯ 
 ͳͲͷِ৘ใ͕֦ࢄͨ͠ɻ 
 ɾ૯຿লͷௐࠪͰɺ͜ΕΒͷِ৘ใͷ͏ͪ 
 ҰͭͰ΋৴ͨ͡ͱ౴͑ͨਓ͸ ೥ྸ૚͕௿͘ͳΔ΄Ͳଟ͘ɺ 
 ̍̑ʙ̍̕ࡀͰ͸̏̒ʹ΋ٴΜͰ͍Δɻ
    Ҿ༻ɿIUUQTXXXTPVNVHPKQNBJO@DPOUFOUQEG
    ʢ৽ܕίϩφ΢Πϧεײછ঱ʹؔ͢Δ৘ใྲྀ௨ௐࠪʣ
11. 
    Ҿ༻ɿIUUQTXXXOBUVSFDPNBSUJDMFTT
    ʢ5IF
    POMJOF
    DPNQFUJUJPO
    CFUXFFO
    QSP
    BOE
    BOUJWBDDJOBUJPO
    WJFXTʣ ϫΫνϯ൓ର೿͕͜ͷ··૿͑ଓ͚Δͱ͍ۙকདྷऔΓࠐ·Εͯ͠·͏ةݥੑ͕͋Δͱ͍͏ݚڀ݁Ռ😱
12. ͍ۙকདྷ΁ͷةػײ ৘ใੜଶܥͰੜ੒͞ΕΔ৘ใ͕૿େ͠ɺ 
 ͔ͭෆ͔֬ͳ৘ใͷׂ߹͕ߴ͍ʢͭ·ΓϊΠζ͕ ଟ͍ʣ৘ใաଟͷঢ়ଶ͕ੜ͡Δ͜ͱͰɺِ৘ใΛ ࣄ࣮ͱޡೝͨ͠Γɺ͋Δ͍͸ࣄ࣮Λࣄ࣮ͩͱ৴͡ ΒΕͳ͍Α͏ͳ͜ͱ͕සൟʹى͜ΔΑ͏ʹͳΕ ͹ɺ೔ৗੜ׆΍ܦࡁ׆ಈɺ͞Βʹ͸ຽओओٛʹ΋ ෛͷӨڹ͕ٴͿՄೳੑ͕͋Δɻ
13. ʁ Έͳ͞ΜɺϑΣΠΫχϡʔε
    ରࡦ͸΍ͬͯ·͔͢ʁ ʁ
14.   ͍͍͑ ͸͍ 
    ΠϯελάϥϜʹͯνʔϜϝϯόʔௐ΂ ʮϑΣΠΫχϡʔεʹԿ͔͠ΒରࡦΛ͍ͯ͠Δ͔ʯ ߴߍੜɾେֶੜਓ
15. ͔͠͠ϑΣΠΫχϡʔε ʹର͢ΔҙࣝΛ͍࣋ͬͯ Δਓ͸·ͩ·ͩগͳ͍ ௐࠪͷ݁Ռ ϑΣΠΫχϡʔεʹ৮ΕΔස౓͸೔ʹ೔ʹ૿͍͑ͯΔ
16.  ײ৘΍ओ؍ʹࠨӈ͞Εͣ
    ٬؍తͳࠜڌ
    Λ࣋ͬͯ΋Β͍͍ͨ Ծઆɾ໨ඪ ؆୯ʹϑΣΠΫχϡʔε൑ఆʹ৮ΕΒΕΔ 
 ؀ڥ͕ඞཁͰ͋ΔͷͰ͸ͳ͍͔ɻ 
17. ੍࡞෺ ✌੍࡞෺
18. IUUQTUXJUUFSDPNUSFOE@PXM
19. σβΠϯͷϙΠϯτ ը૾ʹ਺஋Λॻ͔ͳ͍ ஫ҙࣄ߲ΛಡΜͰ΋Β͏ͨΊ ʹɺ͋͑ͯςΩετʹ͔݁͠ ՌΛॻ͖·ͤΜͰͨ͠ πΠʔτΛݟΔͱ͖ʹ
    ਓؒ͸ը૾͔ΒݟΔ 👎
20. ੍࡞෺ 👩🎓෼ੳख๏
21. ఻೻Ϟσϧºػցֶश Ҿ༻ɿIUUQTXXXSJLFMBCKQTUVEZ ਖ਼͍͠৘ใ '",&
22. IUUQTXXXBBBJPSHPDTJOEFYQIQ"""*"""*QBQFSWJFX'JMF ൑அख๏ ɾΞΧ΢ϯτ͕ొ࿥͞Ε͔ͯΒͲΕ͙Β͍͔
    ɾݩπΠʔτ͔πΠʔτ͞Ε͔ͯΒͲΕ͙Β͍ܦա͍ͯ͠Δ͔ 
 ɾೝূࡁΈΞΧ΢ϯτ͔Ͳ͏͔ 
 ɾϓϩϑΟʔϧը૾Λࢦఆ͍ͯ͠Δ͔Ͳ͏͔ 
 ɾ͍͍Ͷ͍ͯ͠Δ਺ 
 ɾϑΥϩϫʔͷ਺ 
 ɾϑΥϩʔ͍ͯ͠Δ਺ 
 ɾ໊લͷจࣈ਺ 
 ɾࣗݾ঺հͷจࣈ਺ πΠʔτจষʹҰ੾ͱΒΘΕͣ෼ੳΛߦ͏ 📌 ࢀߟ࿦จ
23. ੍࡞෺ ✍Ϟσϧͷਫ਼౓ɾಁ໌ੑ
24. ֶशϞσϧ͸ਫ਼౓௒͑ Ϟσϧͷ༧ଌ ࣮
    ࡍ    
25. ಁ໌ੑϨϙʔτΛ࡞੒ެ։ tensorboardXΛ༻͍ͨϞσϧͷՄࢹԽ
26. ෼ੳ 🧠 πΠʔτऩू IUUQTUXJUUFSDPNUSFOE@PXM ϑΣΠΫ෼ੳ 🤖 ใࠂ
27. IUUQTUSFOEPXMTVHPLVOBSJUBJEFW ϙʔλϧαΠτ 
    աڈͷ෼ੳ͕ݟΕΔ 
    ෼ੳϦΫΤετ͕ૹΕΔ
28. "NB[PO
    $MPVE8BUDI
    &WFOU πΠʔτ "NB[PO
    4 "NB[PO
    -BNCEB "NB[PO
    %ZOBNP%# ఆظ࣮ߦ ෼ੳ ؅ཧը໘ ϩά ը૾Ξοϓϩʔυ ٕज़ߏ੒ (PPHMF
    "QQ
    4DSJQU ϑϩϯτΤϯυ
29. ੍࡞෺ 😁5SFOE0XM΁ͷظ଴
30. ϑΣΠΫ χ ϡ ʔε ͸ ૣ ͘ ޿ ͕ Δ
31. ιʔείʔυ΍Ϟσϧɺ஌ݟͷެ։ 
 ڵຯΛ࣋ͬͯͩ͘͞ΔํΛ૿΍͠ 
 ք۾શମͷٕज़ྗΛ্͛Δ ࠓޙͷల๬ ҙݟΛ΋Β͏ɾ޿ΊΔ ݚڀػؔ΍اۀͱͷڞಉݚڀ 
 ༷ʑੈ୅΍૚ʹ֦ࢄ ެ։ɾڞ༗͢Δ ਫ਼౓ΛߴΊΔ ࣗવݴޠॲཧͱ఻೻Ϟσϧͷྑ఺Λ
    ༻͍ͨϞσϧͷ։ൃ ΞϓϩʔνΛ૿΍͢ 5XJUUFS΍8FCҎ֎ͰͷϦϦʔε 
 ༷ʑͳΞόλʔઃఆͰݕূΛߦ͏
33. ༧උࢿྉ
34. ϑϩϯτΤϯυ Ϧʔμʔ ػցֶशɾαʔόʔαΠυ Α͏͔Μ ͿΜͿΜ ͍ʔͪΌΜ 5ZMFS Ωϯδϣ΢ ͱΈͬ͘͢ αʔϕΠɾσβΠϯ αʔϕΠɾϢʔβΠϯλϏϡʔ αʔόʔαΠυ
35. 4DSBQCPY νʔϜ࿈ܞ /PUJPO 4MBDL
36. ͳͥɺʮϑΝΫτνΣοΫʯͱ͍͏ 
 ΞϓϩʔνΛऔ͍ͬͯΔͷ͔ʁ
37. ৘ใੜଶܥ͸ਐԽ͠ଓ͚͍ͯΔ
    Ҿ༻ɿIUUQTXXXEFMPJUUFDPNKQKBQBHFTTUSBUFHZBSUJDMFTDCTJOGPSNBUJPOFQJEFNJDIUNM 
 ʢੈلͰສഒʹ૿େͨ͠৘ใ఻ୡྗʙ৘ใͷٸ଎ͳ఻છʮΠϯϑΥσϛοΫʯͱ͸ʣ
39. օ͞ΜͷྗΛି͍ͯͩ͘͠͞ 5XJ UU FS ͷதͷਓͱ͓஌Γ߹͍ͷํ
    ޙ΄Ͳ͓ܨ͍͖͍͗ͨͩͨͰ͢ʂ
40. ࢲୡ͕΍͍ͬͯΔ͜ͱ͸Կ͕೔ຊൃɺੈքॳͷϓϩμΫτͳͷ͔
    ैདྷ΍ڝ߹ͱԿ͕ҧ͏ͷ طଘͷ࢓૊Έͱͷҧ͍ /-1 ࣗવݴޠॲཧ Ͱ
    Ͱ͖Δ͜ͱ طଘͷࣙॻɾίʔύεΛ΋ͱʹ
    จ຺Λղੳ͢Δ ٯʹݴ͑͹ ৽͍͠෺ࣄɾະ஌ͷ໰୊ʹ
    ରॲͰ͖ͳ͍͔΋͠Εͳ͍
41. ࢲୡ͕΍͍ͬͯΔ͜ͱ͸Կ͕೔ຊൃɺੈքॳͷϓϩμΫτͳͷ͔
    ैདྷ΍ڝ߹ͱԿ͕ҧ͏ͷ طଘͷ࢓૊Έͱͷҧ͍ ఻೻Ϟσϧ͸ղܾ͠·͢ɻ ະ஌ͷࣄ࣮ʹରͯ͠΋
    ٬؍తͳ෼ੳ͕Մೳ
42. ࢲୡ͕΍͍ͬͯΔ͜ͱ͸Կ͕೔ຊൃɺੈքॳͷϓϩμΫτͳͷ͔
    ैདྷ΍ڝ߹ͱԿ͕ҧ͏ͷ ػցֶशϞσϧͷ࢓૊Έ 3// -45.
    ࠶ؼܕ
    χϡʔϥϧ
    ωοτϫʔΫ
    ϩʔΧϧͳ෼ੳ $//
    ৞ΈࠐΈ
    χϡʔϥϧ
    ωοτϫʔΫ
    άϩʔόϧͳ෼ੳ
    /-1Ͱকདྷతʹਫ਼౓޲্Λʂ [ࢀߟจݙ] Yang Liu, Yi-Fang Brook Wu (2018) Early Detection of Fake News on Social Media Through Propagation Path Classi fi cation with Recurrent and Convolutional Networks