Ntụziaka ntọala

Ịkpọ onwe onye mma-emegharị

Ndozi egwuregwu nke onwe ya na-eme ka ihe atụ dịkwuo mma site n'ime ka ọ na-asọ mpi ma ọ bụ mụta ihe site na nsonaazụ ya gara aga, na-emepụta akara ọzụzụ nke ya.

2 nkeji na-agụEmelitere ikpeazụ

Nchịkọta

Ọ dị mkpa n'ihi na ọ nwere ike ịkwaga arụmọrụ karịa data a na-elekọta na-eji obere akara ma ọ bụ enweghị mgbakwunye mmadụ.

Ime miri emi

Egwuregwu onwe ya nwere mgbọrọgwụ miri emi na egwuregwu AI: AlphaGo Zero na AlphaZero ruru egwuregwu karịrị mmadụ naanị site n'ịkpọ ọtụtụ nde egwuregwu megide onwe ha, na-enweghị ndekọ egwuregwu mmadụ. Otu mmụọ ahụ na-apụta ugbu a na nlegharị anya n'ụdị asụsụ. Na SPIN (Self-Play fine-tuNing), ihe nlereanya dị ugbu a na-ewepụta nzaghachi maka mkpali, na ọzụzụ na-akpali ihe nlereanya ahụ ịmata azịza nke ya sitere na nke mbụ mmadụ dere, na-ewere onwe ya dị ka ma onye ọkpụkpọ na onye mmegide. N'ime ugboro ugboro, 'onye mmegide' (ebe nlele gara aga) na-esiwanye ike, yabụ ihe nlereanya ahụ ga-aga n'ihu na-akawanyewanyewanye, jiri nwayọọ nwayọọ mechie oghere ahụ na nkesa ebumnuche. Nnukwu mkpesa bụ arụmọrụ data: enwere ike ịpịkọta dataset echekwara maka uru ndị ọzọ na-anakọtaghị ngosi ọhụrụ ma ọ bụ mmasị mmadụ.

Nghọta nka nka

SPIN na-emezigharị nke ọma dị ka egwuregwu ọkpụkpọ abụọ nwere mfu ụdị DPO: a zụrụ ihe nlereanya ahụ inye ohere dị elu na nzaghachi ntụaka mmadụ karịa nke onwe ya mepụtara site na mbido mbụ. N'ihi na ebe nlele gara aga na-enye ihe ndị na-adịghị mma, ihe isi ike na-agbanye na-akpaghị aka ka ihe nlereanya ahụ na-akawanye mma. Na usoro egwuregwu, egwuregwu nke onwe na-ejikọta na ọchụchọ (dịka, MCTS) na netwọk bara uru, na-emepụta usoro ọmụmụ na-adịghị agwụ agwụ nke ndị mmegide siri ike na-aga n'ihu na-enweghị data mpụga.

Mmetụta atụmatụ

Mkpebi doro anya

Ọ na-enyere gị aka ikewapụta nkwupụta ọrụ aka doro anya na asụsụ ahịa.

Ọnụ ego na mmefu ego

Ị nwere ike ịjụ ajụjụ mmejuputa iwu ka mma tupu itinye ego ma ọ bụ oge.

Team na usoro ọrụ

Ndị otu nwere nghọta na-eme ka ngwaahịa, amụma na mkpebi mmụta ka mma.

Ọdịnihu nke ịme egwuregwu onwe onye dị mma

Egwuregwu onwe ya bụ onye na-eduga n'ịkwatu mgbidi data, ebe ọ bụ na ọ na-emepụta usoro ọmụmụ nke ya kama ịdabere na aha mmadụ dị ụkọ. Na-atụ anya ito eto na ngalaba ndị enwere ike ịnwapụta dị ka mgbakọ na mwepụ, koodu, na theorem na-egosi, ebe ndị na-enyocha ihe na-akpaghị aka na-enweta ọkwa nke onwe ha. Ihe ize ndụ gụnyere nkwụghachi ụgwọ nkwụghachi ụgwọ na ihe nlere anya daa site na ọzụzụ maka oke mmepụta sịntetik, yabụ sistemụ n'ọdịnihu ga-agwakọta egwuregwu onwe ya na akara mgbada, ihe nrịbama, na nzaghachi mmadụ ma ọ bụ nke ụwa n'ezie.

Mmejuputa n'ezie n'ụwa

AlphaGo Zero na AlphaZero na-erute Go, Chess na shogi zuru oke site na ịme onwe ya n'enweghị egwuregwu mmadụ.

SPIN na-ebuli akara akara LLM site n'ịkọpụta ihe dị iche iche nke ya na azịza ntụaka mmadụ.

Ụdị mgbakọ na mwepụ na koodu ntinye na-emepụta mbọ ngwọta, wee na-azụ ọzụzụ na ndị nyocha akpaka ma ọ bụ nyocha otu

Ndị ọrụ mkparita ụka na mkparịta ụka na-emeziwanye atụmatụ site n'ịkpọ akụkụ abụọ nke mkparịta ụka ugboro ugboro megide onwe ha

Ihe ize ndụ & okporo ụzọ nche

Otu dị iche iche nwere ike iji otu okwu ahụ mee ihe n'ụzọ dị iche, yabụ kọwapụta oge n'oge.

Ihe nrịbama nwere ike ịdị ike ebe arụmọrụ ụwa na-adaghị adaba.

Ileghara ogo data na atụmatụ nyocha anya na-emepụtakarị nsonaazụ na-adịghị mma.

Map mmejuputa

1

Malite na nkọwa asụsụ dị larịị nke nsonaazụ ịchọrọ.

2

Họrọ otu metrik ịga nke ọma na otu ọnọdụ ọdịda tupu nnwale.

3

Gbaa obere onye na-anya ụgbọ elu nwere data nnọchite anya, ọ bụghị ihe ngosi ngosi na-egbu maramara.

4

Detuo ebe Play-Play Fine-Tuning na-enyere aka yana ebe ụzọ ndị dị mfe dị mma.

Nọgide na-eme nchọpụta

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 Self-Play Fine-Tuning quiz

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

Malite ajụjụ

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

Ajụjụ a na-ajụkarị

Gịnị bụ Self-Play Fine-Tuning?

Ndozi egwuregwu nke onwe ya na-eme ka ihe atụ dịkwuo mma site n'ime ka ọ na-asọ mpi ma ọ bụ mụta ihe site na nsonaazụ ya gara aga, na-emepụta akara ọzụzụ nke ya. Ọ dị mkpa n'ihi na ọ nwere ike ịkwaga arụmọrụ karịa data a na-elekọta na-eji obere akara ma ọ bụ enweghị mgbakwunye mmadụ.

Kedu usoro ama ama ruru egwuregwu Go nke karịrị mmadụ na-eji naanị egwuregwu onwe ya na enweghị ndekọ egwuregwu mmadụ?

AlphaGo Zero mụtara kpamkpam site na egwuregwu ọ na-egwu megide onwe ya, malite na egwuregwu na-enweghị usoro na karịa ụdị ndị a zụrụ azụ na egwuregwu mmadụ.

Na SPIN, gịnị na-arụ ọrụ nke 'onye mmegide' nke ihe nlereanya ahụ ga-akụrịrị?

SPIN na-ewere nlegharị anya nke ọma dị ka egwuregwu ịme onwe ya ebe ihe nrụpụta onwe nke izizi izizi na-arụ ọrụ dị ka ihe na-adịghị mma nke ihe nlereanya na-amụta ka ọ dị elu.

Gịnị bụ isi uru data-arụmọrụ nke onwe-egwu onwe-egwu ọma-nlegharị anya?

Egwuregwu onwe ya na-arụpụta mgbama ọzụzụ nke ya, yabụ nhazi a na-ahụ maka ya nwere ike iwepụta nkwalite ndị ọzọ na-anakọtaghị ngosi ọhụrụ.

N'ebumnobi ọzụzụ SPIN, gịnị ka a na-akwali ihe nlereanya ime?

SPIN na-eji mfu ụdị DPO nke na-enye ụgwọ ọrụ ịwapụta azịza ntụaka mmadụ (nke ọma) site na azịza nke onwe onye mbụ mepụtara (adịghị mma).

Kedu ihe kpatara ihe isi ike na-egwu onwe onye na-emegharị nke ọma na-abawanye na-akpaghị aka karịa ugboro ugboro?

N'ihi na ihe na-adịghị mma nke ọ bụla sitere na ụdị siri ike nke gara aga, ihe nlereanya a na-eche ihe ịma aka siri ike na-aga nke ọma na-enweghị nhazi usoro ọmụmụ akwụkwọ ntuziaka.