nyoka-dqn lokola nzela ya kokende na ML

Программирование

Ezwami na tensorflow-js oyo osengeli na yango mpo na ba étapes ya liboso na ML mpe na mombongo.

Oyo ezali dqn formation esika ba recompenses na ba pénalités epesami na agent pona ba actions.

https://github.com/tensorflow/bandakisa-ya-tfjs/nzete/nkolo/nyoka-dqn

https://bobombi.googleapis.com/ndakisa-ya-tfjs/nyoka-dqn/index.html

Il se révèle que tozali na nioka oyo ezuaka liteya na expérience na yango moko pona ko bouger. Azali na nombre moko boye ya ba mouvements (kobaluka to ko continuer) mpe mosala na ye ezali ya kozua mbuma. Na ebandeli, nyoka eyebaka te ndenge nini, kasi na boumeli ya formasyo, ekómaka na likoki ya koluka mbuma mpe kokima. Ezali mpenza ndenge moko na mombongo, boye te?

Bon, ata moke te mpo na mosala na biso, ntango tozali na ba données oyo ezali kotambola, mpe robot esengeli kozwa ekateli ya kosomba to koteka.

Moins ya nioka ezali ete ekomami mpo na node.js, mpe tosengeli kozala na makoki ya kopesa formasyo na navigateur (mpo ete mosaleli nyonso akoki kosala yango). Yango wana, esengelaki mpe kosala bongo mpo na client.

Na bendaki nyoka ndenge ezali na page na ngai na tableau mpe namekaki ko lancer yango. Ezalaki wana te!

Nyoka yango ebundisaki yango

ba paquets liboso, .

na nsima, ba arguments. (par ailleurs, nasalaki pe demande ya kobenda pembeni ya fixation https://github.com/tensorflow/tfjs-examples/pull/353),

sima %%Boboyi oyo esimbami te (TypeError): Ekoki te kobenga kelasi lokola fonction%%. Oyo ezali kobongisama awa https://github.com/tensorflow/tfjs/pull/3906/fichiers, kasi banda Libération ezali nanu te, yango wana esengelaki na substituer yango na maboko. Awa, par ailleurs, batu mingi baza na problème https://github.com/tensorflow/tfjs/issues/3384.

Ba mineurs esi ekende mosika, lokola ko remplacer mosala depuis fs na indexeddb. Mosala oyo na espionaki awa, par ailleurs. Bon, en général, cart-pole na ebandeli e inspiraki ngai nasala nionso na client.

https://github.com/tensorflow/bandakisa-ya-tfjs/nzete/nkolo/pôle-ya-chariot

https://bobombi.googleapis.com/ndakisa-ya-tfjs/pôle-ya-chariot/dist/index.html

na nsuka, oyo nde esalemaki:

https://github.com/pskucherov/opexflow/kobenda/16/ba fisyé

Bon, ya solo, likambo oyo ekolanda ekozala:

  1. Ebongi na ba chartes
  2. Komona makambo na makanisi
  3. Former robot na ba données oyo eyebani pona kolakisa résultat maximum

Awa na se ezali démonstration vidéo ya travail-formation ya ba robots na client. Mokolo mosusu bako trader pe ba futures ya Sberbank au lieu ya ngai.

https://youtu.be/fvCKz8n5whE ezalaka ndenge nini?
pskucherov
Rate author
Add a comment