Upgrade to Pro
— share decks privately, control downloads, hide ads and more …
Speaker Deck
Features
Speaker Deck
PRO
Sign in
Sign up for free
Search
Search
Lasso.pdf
Search
kichinosukey
December 26, 2018
0
51
Lasso.pdf
スパーシティ大事。以下の本の理解を残したくて。著者の方には感謝しかないです。
https://www.sbcr.jp/products/4797393965.html
kichinosukey
December 26, 2018
Tweet
Share
More Decks by kichinosukey
See All by kichinosukey
カーネル関数とは
kichinosukey
0
49
Latin Hypercube Sampling
kichinosukey
0
69
NSGAII
kichinosukey
0
45
NSGA
kichinosukey
0
50
基底関数回帰の概要について
kichinosukey
0
72
児童虐待の現状
kichinosukey
0
56
イノベーションのジレンマとフィンテック
kichinosukey
0
69
Paxos made simpleを理解しよう
kichinosukey
0
420
Featured
See All Featured
Refactoring Trust on Your Teams (GOTO; Chicago 2020)
rmw
32
2.7k
Gamification - CAS2011
davidbonilla
80
5.1k
Java REST API Framework Comparison - PWX 2021
mraible
28
8.3k
Statistics for Hackers
jakevdp
796
220k
For a Future-Friendly Web
brad_frost
175
9.4k
The Web Performance Landscape in 2024 [PerfNow 2024]
tammyeverts
2
290
How to train your dragon (web standard)
notwaldorf
88
5.7k
GraphQLとの向き合い方2022年版
quramy
44
13k
10 Git Anti Patterns You Should be Aware of
lemiorhan
PRO
656
59k
RailsConf & Balkan Ruby 2019: The Past, Present, and Future of Rails at GitHub
eileencodes
132
33k
The Straight Up "How To Draw Better" Workshop
denniskardys
232
140k
XXLCSS - How to scale CSS and keep your sanity
sugarenia
248
1.3M
Transcript
-BTTP
֓ཁ ↟-ϓϰϥΤᩆ᳀ặḄͺͩ൰͚ͱᤎ᷀ᙚΤᱠᖀặͫΝ φ = 1 2 y − ˜ Xω
+ λ ω1 ω1 = d ∑ i=1 ωi
ղ๏֓ཁ ↟᫉ᶝ෦᷀ď$PPSEJOBUF%FTDFOUĐ ↟ؔȃ Y ΤᱠᖀặͫΝͱΓfᤎ᷀ΤຬͱͫḑΤᄌͱ͖ ↟ͩͩ͝ჯᙉͽ᧶ḑΤᄌΝͥͺီ͖ͩf᧱ͼᰮᲩ͝Λξ τ⑲ύͩᓓᙚΤ᙮͖ͩ͡ ↟᧱ͼᰮᲩ ∂φ ∂xi
= 0, (j = 1,…, d) X(0) = (x(0) 1 , x(0) 2 , …, x(0) d )
ղ๏֓ཁ ↟ᤎ᷀ΤຬͱͫΚ͘ͼYΤཏΓfͯΞΤY? L ͺͫΝ ↟ჯ༷ͽ ↟ᓓᙚͶfᵟዸᐠፖᩔᙚΤYͲͣხͩ͝ḑͥ͘ͺ͖ͩΝ ↟ͶมͲͣხͩ͝ͽͼΞᱠ᧱ḑ࿄ᑾͽͼΝͺ͖͘ວ͚ ∂φ ∂x1
(x1 , x(k) 2 , …, x(k) d ) = 0 ∂φ ∂x2 (x(k+1) 1 , x2 , …, x(k) d ) = 0
ղ๏ ↟ͩͩ͝fϮψςȃዸᐠᏵ൱ၙďθϮϥࢀ᪑Đ ↟ͯΐΐ$%͞ᑿ͚ͼ͖ ↟ᠻf᫈ͽᐠᷡͩᵟዸᐠΤḝ͘
ίϥϜ ↟Z]Y]ዸᐠ͟ͼ͖Ħ ↟ᤎ᷀ᙚΚΜᅻଓᕥຬͱͧΞΝ ↟ͩͩ͝ᤎ᷀ΚΜྵᲩ᪰ͩ͞ͼ͖ͱΓͯΐΐዸᐠ͟ͼ͖ lim n→0 f(h) = f(0) =
0 lim h→+0 |x + h| − |x| h = lim h→+0 (x + h) − x h = 1 lim h→−0 |x + h| − |x| h = lim h→+0 −(x + h) − (−x) h = − 1
ղ๏ৄࡉ ↟ᥦᅪḑ͟͟Ξͼ͖ᬊᙚΤᷡᛔhhh ↟ᱠ᪾᧣ͽX@L? X@L?ዸᐠፖᩔᙚḑX@L᙮Ꮞͺͩᑿ ΡΞΝ ↟X@L? X@L X@L?X@LΤ᧸ᩮͺͫΝ ↟X@L? ͺX@L?Τຬͱͧͼ͖ᦉᷯ
X@L ᙮ͩͼ͖ ↟᙮ἦX@LᲩᤎ᷀ͺͼΝ ˜ ωk = S (∑n i=1 (yi − ω0 − ∑ j≠k xi ωj )xik , λ) ∑n i=1 x2 ik ωk
ղ๏ৄࡉ ↟X@L? ͺͼΝͽ ↟X@L?ͺͼΝͽ ↟ͯͩᤎ᷀ᙉX@L? ͝ͶX@L?͔Ν ωk > λ ωk
< − λ −λ < ωk < λ
ιϑτᮢؔ4 ↟THOᐊ߸ؔͺṺΞΝ ↟THO Y Y Y S(p, q)
= sgn(p) max{0, ∥p∥ − q}
ιϑτᮢؔͷಛ ↟X@L᙮͝Λξτ⑲ύ ↟Ϯψςḑၞ͡͞ͺͼΝᔬͻǹX@Lǹฒ᷀ͽͯΞΛ ͔͞ΝͱΓďਤ੨ᔻĐ
ϋΠύʔύϥϝʔλͷޮՌ