Kwiga gushimangira Inyigisho
Kwiga gushimangira inyigisho (IRL) ihindura RL isanzwe: aho guhabwa igihembo no gushaka politiki, ireba imyitwarire yinzobere kandi igahindura imikorere yigihembo cyihishe ibisobanura.
Incamake
This matters because a recovered reward generalizes to new situations far better than directly copied actions.
Kwibira cyane
Kwiga gushimangira inyigisho birabaza: niyihe ntego umuhanga agomba kuba yarakurikiranye kugirango yitware nkuko bakoze? Urebye imyigaragambyo, IRL igarura ibikorwa byigihembo aho iyo myitwarire isa neza (cyangwa hafi-nziza), hanyuma ikoresha RL isanzwe kugirango ikure politiki. Impamvu ni rusange - ibihembo byize byerekana impamvu inyuma yimyitwarire, bityo umukozi ashobora gukora mubwenge muri leta imyigaragambyo itigeze ikubiyemo, bitandukanye na clon yimyitwarire yigana ibikorwa gusa. Ikibazo ntigisanzwe: imirimo myinshi yo guhemba isobanura imyitwarire imwe, harimo noroheje. Inzira zingenzi zikemura iki kidasobanutse, harimo nuburyo ntarengwa bwo guhitamo ibihembo bituma impuguke iba nziza cyane, hamwe na entropie-entropy IRL, itoranya ibihembo bike-byo kugabura bihuye namakuru.
Ubushishozi
Ikibazo nyamukuru ni ukudasobanuka: ibihembo bya zeru bihoraho bituma politiki yose iba nziza, bityo ibihembo byinshi bitarondoreka bisobanura imyigaragambyo iyo ari yo yose. Ntarengwa-entropy IRL ikemura ibi mugushushanya kwerekana nkuko byakuwe mubisaranganya aho amahirwe ya trayectory akura cyane hamwe nibihembo byose. Ibi bitanga intego yihariye, isobanuwe neza kandi mubisanzwe ikemura impuguke zisakuza, zidatunganye, kubera ko inzira zidasanzwe zakira gusa amahirwe yo hasi ariko atari zero aho guhagarikwa.
Ingaruka z'Ingamba
Igiciro na bije
Ibyemezo byubwubatsi bitwara imikorere nigiciro cyimikorere kumyaka.
Ibyemezo bisobanutse
Ubuhanga bwa tekinike bufasha amakipe guhitamo umurongo ukwiye, ntabwo ari shyashya gusa.
Kugenzura ubuziranenge
Guhitamo neza bya injeniyeri bigabanya ibintu byizewe mubikorwa.
Ejo hazaza h'inyigisho zinyuranye zo Kwiga
IRL igenda ishimangira kwiga ibihembo kugirango ihuze: aho kuba abantu ibihembo byandikishijwe intoki, sisitemu zerekana icyo abantu baha agaciro kubitekerezo no gutanga ibitekerezo. Tegereza amahuza akomeye hamwe no kwiga gushimangira ibitekerezo byabantu no kwiga ukunda, gupima imvugo-yerekana imiterere na robo. Ubushakashatsi burimo gusunika kugarura ibihembo bivuye kuri videwo mbisi no kwitegereza igice, no ku bihembo bigaragara ko birwanya ibihembo-hacking n'ibibazo bidasobanutse byugarije uburyo bw'iki gihe.
Gushyira mu bikorwa Isi
Ibinyabiziga byigenga bivuga ibyifuzo byo gutwara (ubworoherane, umutekano wumutekano) kubashoferi babantu
Imashini ziga intego zakazi kuva mubyerekanwe byabantu kugirango zishyire mubikorwa bishya
Kwerekana abanyamaguru cyangwa inyamanswa mugusubiza intego inyuma yinzira zagaragaye
Ibihembo byerekana ibihembo bya AI, wiga indangagaciro zabantu uhereye kumahitamo yerekanwe
Ingaruka & Kurinda
Gutezimbere igipimo kimwe gishobora guhisha intege nke za sisitemu.
Ibikorwa Remezo no kubungabunga akenshi usanga bidahabwa agaciro.
Icyuho cyumutekano no kwitegereza birashobora kwiyongera uko sisitemu igenda igorana.
Igishushanyo mbonera
Sobanura ubukererwe, ubuziranenge, nigiciro cyibiciro mbere yo kubishyira mubikorwa.
Ibipimo byerekana umutwaro ufatika hamwe namakuru yimiterere.
Gukurikirana ibikoresho kubikosa, drift, ningaruka zabakoresha.
Tegura inzira yo gusubiza ibyabaye mbere yo gupima.
Komeza Ubushakashatsi
Free newsletter
Get the daily AI briefing
Three verified AI stories every weekday morning, written in plain English. Free forever, no ads.
One email each weekday. Unsubscribe in one click. We never sell or share your address.
Test yourself
Take the Inverse Reinforcement Learning quiz
Instant feedback on every answer, and a shareable certificate with a verifiable ID once you pass a course.
Support free AI education. AI Understanding is a 501(c)(3) nonprofit — no ads, no paywall, ever. Make a donation
Ubuyobozi bukurikira
Gushimangira Kwiga Biturutse kubitekerezo byabantu
Ibibazo bikunze kubazwa
What is Inverse Reinforcement Learning?
Kwiga gushimangira inyigisho (IRL) ihindura RL isanzwe: aho guhabwa igihembo no gushaka politiki, ireba imyitwarire yinzobere kandi igahindura imikorere yigihembo cyihishe ibisobanura. Ibi bifite akamaro kuko ibihembo byagarutsweho mubisanzwe mubintu bishya neza kuruta ibikorwa byimuwe.
Niki kwigira imbaraga zinyuranye bigamije gukira?
IRL ifata imyiyerekano nkiyinjiza kandi igashiramo ibikorwa byigihembo aho imyitwarire yinzobere igaragara neza.
Kuki ikibazo cya IRL gisobanurwa nkicyiza cyangwa kidasobanutse?
Ibikorwa byinshi byo guhemba, harimo ibihembo bito-byose bya zeru, birashobora gutuma imyitwarire igaragara neza, bityo ibihembo ntabwo bigenwa gusa nimyiyerekano yonyine.
Ni izihe nyungu zingenzi kugarura ibihembo bifite kurenza cloni yimyitwarire?
Igikorwa cyo guhemba gikubiyemo impamvu impuguke yakoze nkuko yabikoze, ikareka politiki ikomokaho ikitwara neza muri leta nshya, mugihe cloni ikoporora ibikorwa bigaragara mumibare.
Nigute ntarengwa-entropy IRL ikemura ibihembo bidasobanutse?
Max-entropy IRL itekereza ko ibihembo byinshi bihebuje birashoboka cyane, bitanga intego idasanzwe nayo yihanganira impuguke zidatunganye, zisakuza.
IRL imaze gukira ibikorwa byo guhemba, mubisanzwe bikorwa iki gikurikira?
IRL itanga ibihembo, hanyuma igashyikirizwa algorithm isanzwe ya RL kugirango ibare politiki nziza muri iyo ntego.