[DL輪読会]最新の深層強化学習

>100 Views

February 17, 17

スライド概要

2017/2/17
Deep Learning JP:
http://deeplearning.jp/seminar-2/

シェア

またはPlayer版

埋め込む »CMSなどでJSが使えない場合

関連スライド

各ページのテキスト
1.

%-IBDL ࠷৽ͷਂ૚‫ڧ‬Խֶश ૑଄৘ใֶઐ߈. ॳ୩ྯ࣊

2.

#BDLHSPVOE w ਂ૚‫ڧ‬ԽֶशͰৗউ'9USBEFSΛҭ͍ͯͨ

3.

0VUMJOF w ਂ૚‫ڧ‬Խֶश͓͞Β͍ w "$&3 w 6/3&"- w 21SPQ w 1(2

4.

ਂ૚‫ڧ‬Խֶश͓͞Β͍

5.

%2/ w WBMVFCBTFE ํࡦΛཅʹද‫ ͍ͳ͠ݱ‬ w ߦಈՁ஋ؔ਺ 2ؔ਺ Λ//Ͱද‫ݱ‬ w P⒎QPMJDZͰFYQFSJFODFSFQMBZΛ࢖ͬͯߋ৽ w ޯ഑ʹۙࣅ2Λ࢖͍ͬͯΔͨΊCJBTFE 1MBZJOH"UBSJXJUI%FFQ3FJOGPSDFNFOU-FBSOJOH  .OJIFUBM

6.

"$ w QPMJDZͱঢ়ଶՁ஋ؔ਺7Λ//Ͱද‫ݱ‬ w POQPMJDZ ෳ਺"HFOUͰඇಉ‫ֶظ‬श w POQPMJDZͷͨΊલͷQPMJDZͷαϯϓϧ͸࢖͑ͳ͍ w ਤதLͰCJBTWBSJBODFUSBEFP⒎ΛDPOUSPM " "TZODISPOPVT.FUIPETGPS%FFQ3FJOGPSDFNFOU-FBSOJOH  .OJIFUBM

7.

%%1( w QPMJDZΛֶश w QPMJDZΛ֬཰෼෍Ͱͳܾ͘ఆతͳؔ਺ͱͯ͠ද‫ݱ‬ w P⒎QPMJDZͰTBNQMFF⒏DJFOU w ޯ഑ʹۙࣅ2Λ࢖͍ͬͯΔͨΊCJBTFE $POUJOVPVTDPOUSPMXJUIEFFQSFJOGPSDFNFOUMFBSOJOH  -JMMJDSBQFUBM

8.

େମͷτϨϯυ ओ‫؍‬ w w w WBMVFCBTFEΑΓQPMJDZCBTFE w ࿈ଓ஋Λද‫͖Ͱݱ‬Δ w ํࡦΛ֬཰తʹද‫͖Ͱݱ‬Δ POQPMJDZΑΓP⒎QPMJDZ w ‫ੈ࣮ݱ‬ք΁ͷԠ༻࣌ʹPOQPMJDZͩͱίετߴ TBNQMFF⒏DJFOU  w ํࡦ͕‫ॴہ‬ղʹؕΓͮΒ͍ CJBTWBSJBODFUSBEFP⒎Λվળͯ͠ऩଋΛ଎͘΋͘͠͸҆ఆͤ͞Δ w "$͸POQPMJDZ͔͠࢖͑ͳ͍ w %%1(͸P⒎QPMJDZ࢖͑Δ͚Ͳ2ͷޯ഑࢖͏͔ΒCJBTߴͰ҆ఆ͠ͳ͍

9.

7BMVFCBTFE 1PMJDZCBTFE POQPMJDZ WBSJBODFߴ "$ %2/ P⒎QPMJDZ CJBTߴ %%1(

10.

21SPQ 21SPQ4BNQMF&⒏DJFOU1PMJDZ(SBEJFOUXJUI"O0⒎1PMJDZ$SJUJD  (VFUBM

11.

21SPQᶃ w ςʔϥʔల։ͷ߲̍࣍Λ࢖ͬͯMJLFMJIPPESBUJPʹ %%1(ͷޯ഑Λ௥Ճ w ௥Ճ߲ʹ܎਺Λಋೖͯ͠CJBTҰఆͷ··WBSJBODFΛ Լ͛Δ w %%1(ͷ߲Ͱ࢖͏2͸P⒎QPMJDZͰߋ৽Մೳ

12.

21SPQᶄ ೚ҙͷؔ਺Gʹ͍ͭͯ੒ཱ ৄࡉ͸࿦จ"QQFOEJY Gʹ"Λ࢖͍ ςʔϥʔల։ΛЖपΓʹ͢Δ MJLFMJIPPESBUJP %%1( ΠʔλΛಋೖͯ͠όΠΞεΛ૿Ճͤ͞ͳ͍·· όϦΞϯεΛ‫ݮ‬গͤ͞Δ

13.

21SPQᶅ w w //Ͱද͢ͷ͸ w QPMJDZɹɹɹɹɹɹɹɹɹɹɹɹɹɹɹɹɹɹɹɹ ЖͱЄͰΨ΢γΞϯQPMJDZ Ж͸%%1(߲Ͱ΋࢖͏  w 7 "EWBOUBHFGVODUJPOͷͨΊ  w 2 %%1(ͷͨΊ  ͜ͷ಺2͸P⒎QPMJDZͰֶशՄೳ

14.

21SPQᶆ

15.

21SPQ·ͱΊ w όΠΞε্͛ͣʹόϦΞϯεΛԼ͛ͯ҆ఆੑͷ޲্ w 2ΛP⒎QPMJDZͰֶश w ΠʔλΛ্ख͘બͿͱ2ͷֶश೗ԿʹґΒֶͣश͕ ҆ఆ͢Δ

16.

7BMVFCBTFE 1PMJDZCBTFE POQPMJDZ WBSJBODFߴ "$ 21SPQ %2/ P⒎QPMJDZ CJBTߴ %%1(

17.

6/3&"- 6/TVQFSWJTFE3&JOGPSDFNFOU BOE"VYJMJBSZ-FBSOJOH 3FJOGPSDFNFOU-FBSOJOHXJUI6OTVQFSWJTFE"VYJMJBSZ5BTLT  +BEFSCBSHFUBM

18.

6/3&"-ᶃ w ಉ͡ωοτϫʔΫΛ࢖ͬͯิॿతͳλεΫΛղ͔ͤ Δ͜ͱͰऩଋΛ଎͘͢Δ w ิॿλεΫʹ͸FYQFSJFODFSFQMBZ͕࢖͑Δ w ਓ͕ເΛ‫ݟ‬ΔͷΛ‫͑ߟʹݩ‬ΒΕ͍ͯΔ

19.

6/3&"-ᶄ

20.

6/3&"-ᶅ w SFXBSEQSFEJDUJPOͷ࣌͸SFXBSE͕ͳ͍֬཰͕ ʹͳΔΑ͏ʹ&3͔Βαϯϓϧ͢Δ w QJYFMDPOUSPMΑ͘Θ͔Βͳ͍ w ऩଋ͸"$ΑΓ଎͍͠ੑೳྑ͍ w ϩόετੑ΋͋Δ

21.

7BMVFCBTFE 1PMJDZCBTFE POQPMJDZ WBSJBODFߴ "$ 21SPQ %2/ P⒎QPMJDZ CJBTߴ %%1( 6/3&"-

22.

"$&3 "DUPS$SJUJD&YQFSJFODF3FQMBZ 4BNQMF&⒏DJFOU"DUPS$SJUJDXJUI&YQFSJFODF3FQMBZ  8BOHFUBM

23.

"$&3 w ॏཁ౓αϯϓϦϯάΛߦ͏͜ͱͰP⒎QPMJDZͷσʔ λΛ࢖ͬͯ‫ࡏݱ‬ͷQPMJDZΛֶश w 3FUSBDFΛ࢖ͬͯॏཁ౓ॏΈΛ੍‫ݶ‬ w 5SVTU3FHJPO1PMJDZ0QUJNJ[BUJPOΛޮ཰Խ 4BGFBOEF⒏DJFOUP⒎QPMJDZSFJOGPSDFNFOUMFBSOJOH  .VOPTFUBM

24.

7BMVFCBTFE 1PMJDZCBTFE POQPMJDZ WBSJBODFߴ "$ 6/3&"- 21SPQ %2/ P⒎QPMJDZ CJBTߴ %%1( "$&3

25.

1(2 1(2$PNCJOJOHQPMJDZHSBEJFOUBOE2MFBSOJOH  0`%POPHIVFFUBM

26.

1(2 w ऩଋ‫ݻ‬ఆ఺ʹ͓͍ͯ੒Γཱͭ౳ࣜΛ࢖ͬͯ2ΛQPMJDZ Λ࢖ͬͯද‫ݱ‬ w ͦͷ2ʹରͯ͠2MFBSOJOHΛߦ͏͜ͱͰQPMJDZΛ VQEBUF

27.

1PMJDZCBTFE POQPMJDZ WBSJBODFߴ 7BMVFCBTFE "$ 1(2 6/3&"- 21SPQ %2/ P⒎QPMJDZ CJBTߴ %%1( "$&3

28.

·ͱΊ w "$΍%%1(͕࣋ͭ՝୊͕໌Β͔ʹͳ͖ͬͯͯTBNQMF F⒏DJFODZ΍CJBTWBSJBODFUSBEFP⒎ͷվળ͕ॏཁ w "$&3ͷΑ͏ͳૉ๿ʹP⒎QPMJDZΛద༻͢Δͷ΋͋Ε͹ 6/3&"-ͷΑ͏ʹ‫ڧ‬Խֶशͱ͍͏ΑΓ//ͷಛੑΛ ࢖ͬͨ΋ͷ΋͋Γ‫ڵ‬ຯਂ͍ w ‫ݸ‬ਓతʹ͸21SPQ͕Ұ൪Ͳ͏͍͏ൃ૝Ͱͦ͏ͨ͠ͷ͔ Θ͔Βͳ͔ͬͨ(V͞Μͷ.V1SPQͱ͔ͪΌΜͱಡΊ ͹Θ͔Δ͔΋͠Εͳ͍