GUIDE teknik

Jàng buy toppandoo

Jàngat buy toppandoo dafay jàngal IA def benn liggéey ci kopie li eksper yi wane, moo gën ñuy jàngee ci jéem-ak-njuumte.

2 simili jàngDañu mujjee yeesal

Résumé

It matters because for many real tasks — driving, surgery, manipulation — it is far easier to show good behavior than to write a reward function.

Plongeur bu xóot

Jàngat buy toppandoo dafay tàggat ab politik ci misaal yuñ bind ci ab eksper buy doxal ci ab environmaa, lu ci gëna bari ay seetlu ak jëf yu eksper bi jël. Xeetu klonaas bi gëna yomba def, dafay jàppee lii ni jàng bu ñuy saytu bu leer: wax luy waaja am ci jëfi eksper bi, buñu sukkandikoo ci réew mi. Dafay neex sudee neexal yi jafe nañu leeral waaye ñu bari ñu koy wane, lu melni ci oto yuy dawal seen bopp yuñ tàggat ci nit ñi wala ay robo yuñ jàngal ci teleoperaasioŋ. Neexaay bi gëna mag mooy coppite ci séddale bi, wala njuumte buy gëna yokk: njuumte yu ndaw yi ci wax luy waaja am dañuy push agent bi ci etaa yi expert bi musul dem, fu amul benn njiit ba noppi di dem fu sori. Pexe yu melni DAgger saafara lii ci laaj eksper bi lu bari yoon ci réew yi jàngat bi yegg.

Gis-gis xarala

Clonage behavioral dafay wàññi perte buñ saytu diggante jëf yiñ séentu ak yiñ wane, waaye dafay xalaat ni réew yi moom seen bopp te dañu séddale ci anam wu wuute - njuumte ci seytu yu toppalante. DAgger (Dataset Aggregation) dafay dog ​​xalaat boobu ci di gëna jëmmal ci politik bi fi nekk, laaj eksper bi mu joxe etiketu etaa yi ñu dem, ak tàggataat ci done yiñ dajale di gëna màgg. Loolu dafay tax done yi ñuy tàggat ñu méngoo ak distribusioŋ réewu jàngat bi, di wàññi bu baax njuumte yiy gëna yokk ci diir bu xawa yàgg.

njeextalu pexe

Njëgg ak budget

Dogal yi architecture di jël dañuy indi njariñ ak njëgu liggéey bi ay at ci ginaaw.

dogal yu gëna leer

Njàngalem xarala yi dafay jàppale ekip yi ñu tànn li gën, te baña yam ci li gëna bees daal.

Xool kalite

Tanneef yu gëna baax ci wàllu ingeñër dina wàññi jafe-jafe yi ci wàllu wóor ci liggéey bi.

Ëlëgu jàng buy toppandoo

Jàngale ci toppandoo lu am solo la ci yokkuteg modelu fondasioŋu robot, fu benn politik di tàggat ci ay done yu bari yu ñuy jëfandikoo ci teleoperaasioŋ, ba noppi ñu defar ko ngir am xam-xam bu bees. Xaarandil fusion bu gëna dëgër ak làkk ak gis-gis suko defee robot yi topp ci wideo yi wala tegtal yi, boole ci ibrid yuy bootstrap ak klonaas ba noppi setal jaaraleko ci jàng buy dooleel. Yokkateg dajale wane ci anam wu yomb, jaaraleko ci simulation ak done yu nit ñi dajale, mingi wéy di nekk bottleneck bu mag bi ak frontiere biy dox.

Doxal ci àdduna dëgg

Modèle oto yuy dawal seen bopp yuñ tàggat ci ni nit ñi di dawal

Ay loxo robot ñu ngi jàng plie ay yéere wala dajale ay mbir ci ay wane yuñ def ci tele

Agent yiy jouer jeu yi dañu leen jëlee ci replay yu nit ñi laata ñuy fine-tuning ak RL

Robot chirurgikaal ak assistant ñu ngi jàng yëngu ci operatër yu xarañ yi

Risk yi ak balustrade yi

Optimize benn benchmark mën na nëbb ñakk kattan yu gëna yaatu ci sistem bi.

Njëg li ñuy fay ci infrastructure yi ak ci toppatoo dañuy faral di suufeel.

Bu sistem yi di gëna xawa jafee xam, jafe-jafe yi am ci wàllu kaaraange ak seetlu mën nañu gëna bari.

Roadmap ngir samp gi

1

Mandargal latency, kalite, ak njëg yi laata ngay jëfandikoo.

2

Benchmark ci biir sargal ak done yu dëggu.

3

Jumtukaay bi di saytu njuumte yi, derive bi ak njeextalu jëfandikukat bi.

4

Waajal rollback ak yooni tontu ci jafe-jafe yi laata ngay eskale.

Weyal di banneexu

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 Imitation Learning quiz

Instant feedback on every answer, and a shareable certificate with a verifiable ID once you pass a course.

Tambalil quiz

Support free AI education. AI Understanding is a 501(c)(3) nonprofit — no ads, no paywall, ever. Make a donation

Gis bi ci topp

Jàngale buy dooleel ci net bi

Laaj yi ñuy faral di laaj

What is Imitation Learning?

Jàngat buy toppandoo dafay jàngal IA def benn liggéey ci kopie li eksper yi wane, moo gën ñuy jàngee ci jéem-ak-njuumte. Dafa am solo ndax ci liggéey yu bari dëgg - dawal, operaasioŋ, manipulasioŋ - dafa gëna yomba wane jikko ju baax moo gën bind fonction neexal.

Lan mooy xalaat bu mag bi ci ginaaw jàng toppandoo?

Jàngat toppandoo dafay tàggat ndawu liggéey bi mu wanewaat jeffin ju ñu wane ci njuréefi eksper yi, moo gën ñu gis jeffin ju jaare ko ci jéem ak njuumte yu lalu ci neexal.

Ban xeetu jafe-jafe mooy klonaasu jeffin?

Clonage behavioral dafay jàppee wane yi ni ay done yuñ etikete, ba noppi jàng xam luy waaja am ci jëfi eksper bi ci etaa bu nekk buñu seetlu, muy nuru jàng buñ yor.

Ban jafe-jafe moo waral klonaasu doxalin bi baña mëna dem ci jëf yu yàgg yu toppalante?

Jumtukaay yu ndaw ci jëf dañuy puus ndawu liggéey bi ci anam wu eksper bi musul wane; amul benn njiit, njuumte yi dañuy gëna bari, te ndawu liggéey bi dafay dem fu sori ci yoonu eksper bi.

naka la algorithm DAgger di saafaraa ñakk kattan googu?

DAgger dafay genne politik bi fi nekk, am etiketu eksper bi ci etaa yi ñuy sooga dem, ba noppi tàggataat ci done yiñ dajale suko defee done tàggat yi méngoo ak distribution etaa bi jàngat bi boppam.

Lan moo waral jàng toppandoo di gëna taamu jàng dooleel ci liggéey yu melni dawal?

Ci liggéey yu jafee def ci àdduna, bind neexal buy wane jikko ju baax lu jafe la, waaye wane misaali jikko ju baax (journal dawal nit) lu yomb la.