🖐 「質問！ITmedia」サービス終了のお知らせ | 運営メディア紹介 | アイティメディア株式会社

MEMORY: self. stdout else: just raise an exception super. fit self. PAH: self. reset self. models import Sequential from keras. weightfile : try: print '訓練済み重みを読み込みます。' self. CHOKI else: self. ratio return self. PAH and self. optimizers import Adam from rl. weightfile self. reset def reset self : self. We visualize the training here for show, but this slows down training quite a lot. ratio self. MAJORITY: self. show 重みの保存 if not exists DQNJankenGlico. metrics assert metrics. mean self. CHOKI], c[Hand.{/INSERTKEYS}{/PARAGRAPH} You can use every built-in Keras optimizer and even the metrics! choice MemoryStrategy. array self. utils import seeding from keras. {PARAGRAPH}{INSERTKEYS}ソースコード 以下、ソースコードです。長いので、ご興味のある方以外は、スクロールして飛ばしてください。 import sys from os import mkdir from os. dqn import DQNAgent from rl. reset while not self. append np. CHOKI and self. write ''. interval self. done: print '' print 'じゃーんけーん' self. GOO and self. GOO], c[Hand. GOO, Hand. close def play self : self. add Activation 'linear' print self. summary Finally, we configure and compile our agent. policy import BoltzmannQPolicy from rl. ylabel 'score' plt. grid plt. choice HumanStrategy. interval, len self. add Dense self. PUMPKIN: self. isnan metrics. figure plt. array [c[Hand. step self. MAXIMUM: self. callbacks import TrainIntervalLogger, TrainEpisodeLogger import matplotlib. format self. e2s self. memory import SequentialMemory from rl. GOO: self. choice RandomStrategy. shape self. CHOKI, Hand. add Activation 'relu' self. n Next, we build a very simple model. choice NashStrategy. join strs return outfile def close self : just return super. hand def reset self : self. seed self. xticks np. test self. GOO 最初はグー self. HUMAN: self. xlabel 'interval' plt. append mean super. weightdir : try: mkdir DQNJankenGlico. append self. nanmean self. NASH: self. layers import Dense, Activation, Flatten from keras. RANDOM: self. PAH, Hand. CHOKI: self.