E fumanoe ho tensorflow-js seo u se hlokang bakeng sa mehato ea pele ho ML le khoebo.
Ena ke koetliso ea dqn moo meputso le likotlo li fuoang moemeli bakeng sa liketso.
https://github.com/tensorflow/tfjs-examples/tree/master/snake-dqn
https://storage.googleapis.com/tfjs-examples/snake-dqn/index.html
Hoa etsahala hore re na le noha e ithutang phihlelo ea eona ho falla. O na le palo e itseng ea metsamao (thinya kapa tsoela pele) mme mosebetsi oa hae ke ho fihla litholoana. Qalong, noha ha e tsebe joang, empa ha e ntse e koetlisoa, e fumana bokhoni ba ho fumana litholoana le dodge. E tšoana haholo le khoebo, na ha ho joalo?
Hantle, bonyane bakeng sa mosebetsi oa rona, ha re na le data e sebetsang, ‘me roboto e hloka ho etsa qeto ea ho reka kapa ho rekisa.
Ntho e fokolang ea noha ke hore e ngotsoe bakeng sa node.js, ‘me re hloka ho khona ho koetlisa ho sebapali (e le hore mosebedisi leha e le ofe a ka e etsa). Ka hona, se tšoanang se ne se lokela ho etsoa ho moreki.
Ke ile ka hula noha joalokaha e le leqepheng la ka ka chate ‘me ka leka ho e qala. E ne e le sieo!
Noha ea itoanela
liphutheloana pele,
ebe likhang. (ka tsela, ke boetse ke entse kopo ea ho hula haufi le tokiso https://github.com/tensorflow/tfjs-examples/pull/353),
ebe %%Hano e sa Tšoaroeng (TypeError): Ha e khone ho bitsa sehlopha e le mosebetsi%%. E ntseng e lokisoa mona https://github.com/tensorflow/tfjs/pull/3906/files, empa ho tloha Ha ho so lokolloe, kahoo ke ile ka tlameha ho e nkela sebaka ka matsoho a ka. Mona, ka tsela, batho ba bangata ba na le bothata https://github.com/tensorflow/tfjs/issues/3384.
Bana ba se ba tsoetse pele, joalo ka ho tlosa mosebetsi ho tsoa ho fs ka indexeddb. Mosebetsi oo ke ileng ka o hloela mona, ka tsela. Ha e le hantle, ka kakaretso, pele-pele e ile ea ntšusumelletsa ho etsa ntho e ‘ngoe le e’ ngoe ho moreki.
https://github.com/tensorflow/tfjs-examples/tree/master/cart-pole
https://storage.googleapis.com/tfjs-examples/cart-pole/dist/index.html
qetellong, sena ke se etsahetseng:
https://github.com/pskucherov/opexflow/pull/16/files
Ha e le hantle, mohato o latelang e tla ba:
- E lumellana le lichate
- Pono ya maikutlo
- Koetlisa roboto ho data e tsebahalang ho bonts’a sephetho se phahameng
Ka tlase ke setšoantšo sa video sa koetliso ea mosebetsi oa liroboto ho moreki. Ka letsatsi le leng ba tla rekisa le bokamoso ba Sberbank ho fapana le nna.
