InoteveraGaidhi rinotevera
GARCH Volatility Models
Tekinoroji
Nhungamiro yehunyanzvi
Iyo Gaussian musanganiswa modhi (GMM) inomiririra kugovera data semusanganiswa wakayerwa weGaussian zvikamu uye inopa yega yekutarisa mikana yenhengo.
Teerera kugwaro iriAI-yakarondedzerwa · ~3 min
Kusiyana ne-k-zvinoreva, inoenzanisira chikamu covariance uye nhengo yakapfava, asi fungidziro dzayo dzeGaussian, kuverenga kwechikamu uye maitiro ekugadzirisa emunharaunda anoda kuongororwa.
Iyo inogumira GMM inoenzanisira density sehuwandu hweK chikamu densities inoyerwa nekusanganisa zviyero. Chimwe nechimwe chikamu iGaussian ine chirevo chayo uye covariance chimiro chakasarudzwa nemuenzaniso. Uremu haurevi uye hunokwana kune imwechete. Kuti utarise, mutemo weBayes unopa basa: mukana wepashure wekuti chikamu chega chega chakagadzira iyo poindi. Iyo yakaoma cluster label inogona kugadzirwa nekusarudza mutoro mukuru, asi kuita kudaro kunorasa kusavimbika. GMMs inogona kumiririra elliptical masumbu, uye yakazara covariance matrices inotora hukama pakati pezvinhu mukati mechikamu chimwe nechimwe. Diagonal covariance inofungidzira kuti hapana mukati me-chikamu chimiro covariance, yakasungwa covariance inogovera imwe general matrix pane zvikamu, uye spherical covariance inoshandisa scalar musiyano pachikamu. Mamwe mamodhi anochinjika ecovariance anoda mamwe ma paramita uye anogona kuwanda kana data ishoma. Kuyera uye zvikamu zvezvikamu zvine basa nekuti covariance inoyerwa munzvimbo yekupinda. EM inowanzoshandiswa kufungidzira GMM paramita. Inochinjanisa mabasa uye inorema parameter inogadziridza. Nekuti mukana wacho hausi conconvex, kutanga kunogona kukanganisa mhinduro. Kudzokorodza kwakawanda kunoderedza kutsamira pane imwe nzvimbo yekutanga asi kusaratidza optimum yepasirese. Covariance kujairana kunogona kudzikamisa pedyo-imwechete fungidziro; imodeli kana manhamba marongero anofanirwa kutaurwa. K-zvinoreva zvinogona kutariswa pasi pefungidziro dzinodzora seane spherical, akaenzana-saizi maGaussian masumbu ane mabasa akaomarara, asi zvibodzwa zvinoshanda zvinosiyana: k-zvinoreva inoderedza nhambwe dzakaenzana uye GMM inowedzera mukana. GMMs inopa density fungidziro uye akapfava maactions, asi usangoerekana wawana nhamba yechokwadi yemapoka ane chinangwa. Shandisa maitiro akadai seBIC serubatsiro rwekusarudza-modheni uye tarisa mukana wakabatika, kugadzikana uye kushanda kwedomasi. Musanganiswa zvikamu zvemasvomhu zvehuwandu hwakakwana uye hunogona kusaenderana nehuwandu hwepasirese hwakasiyana.
Dataset yekufungidzira ine mapoka maviri anopindirana. GMM yakarongedzerwa inogona kupa imwe poindi mutoro 0.7 kune chimwe chikamu uye 0.3 kune chimwe, kuchengetedza kusavimbika pane kuita basa rakaoma ipapo.
Chisumbu chakareba uye chakarereka. Iyo yakazara covariance GMM inogona kumiririra iyo chimiro, nepo spherical covariance modhi inotora chikamu chimwe nechimwe chine musiyano wakagovaniswa munzira dzese.
Muongorori anokodzera akati wandei ezvikamu zviverengero uye covariance marudzi, anoshandisa akawanda ekutanga uye anoenzanisa nzira yeruzivo uye hunhu hwakabatika pane kusarudza kuverenga kubva pachirongwa chete.
Chikwata chinofananidza mabasa eGMM ane k-zvinoreva mavara. K-zvinoreva inoderedza mukati me-cluster squared madaro, nepo GMM ichifungidzira musanganiswa unoenderana, saka migove inogona kusiyana kunyanya kune anopindirana kana mapoka akasiyana.
Mishumo yeGMM inogona kuvandudza nekubatanidza mikana yenhengo necovariance fungidziro, modhi-yekusarudza humbowo uye kugadzikana kwekutangazve. Zvikwata zvinofanirwa kuongorora kana chikamu chinomiririra pateni inobatsira pane kufunga kuti yega yega Gaussian iboka rechisikigo. Kana zvinocherechedzwa zvasvika nekufamba kwenguva, tarisa mukana uye basa rinochinja kuti uone shanduko yehuwandu. Chirongwa chekusimbisa chinoshanda chinoenzanisa zvimiro zvevamiriri covariance uye chikamu chinoverengerwa pane data risingashandiswe kuenderana navo. Zvirinani kusavimbika kuratidza kunogona kubatsira vashandisi kudzivirira kubata 0.51 mutoro seimwe cluster basa.
Tsamba yemahara
Nhatu dzakasimbiswa dzeAI nyaya mangwanani ega ega pakati pevhiki, dzakanyorwa neChirungu chakajeka. Yemahara zvachose, hapana ads.
Imwe email zuva rega rega revhiki. Bvisa kunyoresa kamwe chete. Hatimbotengese kana kugovera kero yako.
Zviedze iwe pachako
Mhinduro yepakarepo pamhinduro yega yega, uye chitupa chinogovaniswa chine ID inobatika kana wapedza kosi.
Tsigira dzidzo yemahara yeAI. AI Understanding ndeye 501(c)(3) isingabatsiri - hapana ads, hapana paywall, zvachose. Ita chipo
Iyo Gaussian musanganiswa modhi (GMM) inomiririra kugovera data semusanganiswa wakayerwa weGaussian zvikamu uye inopa yega yekutarisa mikana yenhengo. Kusiyana ne-k-zvinoreva, inoenzanisira chikamu covariance uye nhengo yakapfava, asi fungidziro dzayo dzeGaussian, kuverenga kwechikamu uye maitiro ekugadzirisa emunharaunda anoda kuongororwa.
Basa rinopa mukana wakagoverwa kune chimwe nechimwe chikamu chekutarisa ikoko uye chidimbu kune chimwe.
Covariance yakazara inopa chikamu chimwe nechimwe chayo chisina kuganhurirwa covariance matrix, zvichienderana nekunyatsojeka.
A diagonal matrix inoisa off-diagonal covariance mazwi kusvika zero.
Zvinangwa uye humiriri hwenhengo hwakasiyana: kudzikisira chinhambwe nemapoka akaomesesa maringe nezvinobvira zvingangoenderana.
EM inogona kuchinjika kune dzakasiyana mhinduro dzemuno zvichienderana nekutanga paramita.
Ramba uchidzidza
Mamwe madhairekitori akasarudzirwa nyaya iyi
InoteveraGaidhi rinotevera
GARCH Volatility Models
Tekinoroji