Ku kumeka eka tensorflow-js leswi u swi lavaka eka magoza yo sungula eka ML na ku xaviselana.
Lexi i ndzetelo wa dqn laha hakelo na swigwevo swi nyikiwaka muyimeri eka swiendlo.
https://github.com/tensorflow/swikombiso swa-tfjs/murhi/n’wini/nyoka-dqn
https://vuhlayiselo.googleapis.com/swikombiso swa-tfjs/dqn-ya-nyoka/index.html
Swi tikomba leswaku hi na nyoka leyi dyondzaka eka ntokoto wa yona ku famba. Una nhlayo yokarhi ya ku famba (turn or continue) naswona ntirho wa yena iku fika eka mihandzu. Eku sunguleni, nyoka a yi swi tivi leswaku yi njhani, kambe loko yi ri karhi yi leteriwa, yi kuma vuswikoti byo kuma mihandzu ni ku papalata. Swi fana swinene ni ku xaviselana, a swi tano ke?
Kahle kahle, at least eka ntirho wa hina, loko hi ri na data leyi tirhaka, naswona robot yi lava ku endla xiboho xo xava kumbe ku xavisa.
Minus ya nyoka hi leswaku yi tsaleriwe node.js, naswona hi fanele ku kota ku leteriwa eka browser (leswaku mutirhisi wihi na wihi a kota ku swi endla). Hikwalaho, leswi fanaka a swi fanele swi endliwa eka muxavi.
Ndzi kokele nyoka tani hi leswi yi nga xiswona eka pheji ya mina hi chati ndzi ringeta ku yi launch. A swi nga ri kona!
Nyoka yi lwisane na yona
tiphasela ro sungula, .
kutani ku va ni minjhekanjhekisano. (hi ndlela, ndzi tlhele ndzi endla xikombelo xo koka ekusuhi na ku lulamisa https://github.com/tensorflow/tfjs-examples/pull/353),
kutani %%Ku ala loku nga khomiwangiki (TypeError): A swi koteki ku vitana tlilasi tanihi ntirho%%. Lexi xi lunghisiweke laha https://github.com/tensorflow/tfjs/pull/3906/files, kambe ku sukela A ku si humesiwa, kutani ndzi boheke ku yi siva hi mavoko ya mina. Laha, hi ndlela leyi, vanhu vo tala va na xiphiqo https://github.com/tensorflow/tfjs/issues/3384.
Minors se yi ye emahlweni, ku fana na ku siva ntirho ku suka eka fs hi indexeddb. Ntirho lowu ndzi nga hlota ha wona lahawani, hi ndlela leyi. Phela, hi ku angarhela, cart-pole eku sunguleni yi ndzi hlohlotele ku endla hinkwaswo eka muxavi.
https://github.com/tensorflow/swikombiso swa-tfjs/murhi/n’wini/xigolonyana-xigolonyana
https://vuhlayiselo.googleapis.com/swikombiso swa-tfjs/xigolonyana-xigolonyana/dist/index.html
eku heteleleni, leswi hi leswi humeleleke:
https://github.com/pskucherov/opexflow/ku koka/16/tifayela
Phela, entiyisweni, goza leri landzelaka ku ta va:
- Ku ringanela eka tichati
- Ku vona hi mahlo ya mianakanyo
- Dyondzisa robot eka data leyi tivekaka ku kombisa mbuyelo lowukulu
Laha hansi ku na vhidiyo ya demo ya ntirho-ku leteriwa ka tirhoboto eka muxavi. Siku rin’wana va ta tlhela va xaviselana ti futures ta Sberbank ematshan’weni ya mina.
