1. The document summarizes two papers about bandit algorithms. The first paper proposes a multi-level bandit algorithm that utilizes the taxonomy of ads and web pages to reduce the number of arms to explore. The second paper studies the "mortal multi-armed bandit" problem where arms have finite lifetimes. It models the death rates of arms and proposes the "Stochastic with Early Stopping" algorithm that investigates new arms for a fixed number of pulls before abandoning them.
1. The document summarizes two papers about bandit algorithms. The first paper proposes a multi-level bandit algorithm that utilizes the taxonomy of ads and web pages to reduce the number of arms to explore. The second paper studies the "mortal multi-armed bandit" problem where arms have finite lifetimes. It models the death rates of arms and proposes the "Stochastic with Early Stopping" algorithm that investigates new arms for a fixed number of pulls before abandoning them.
6. 2.1.1 タイムゾーン情報の集計(1)
今回の⽬目標:
タイムゾーンのうち、最も頻度度の⾼高いものを調べよう〜~
• リスト内包を使ってタイムゾーンのリストを抽出
time_zones = [rec['tz'] for rec in records]
print time_zones
• だめだった。じゃあ、条件付け加えてなんとかしましょう
# もしこの⾏行行にtzあったら抽出してください〜~ 10⾏行行まで!
time_̲zones = [rec['tz'] for rec in records if 'tz' in rec]
print time_zones[:10]
7. 2.1.1 タイムゾーン情報の集計(2)
• ここまでできたら、タイムゾーンの出現回数をカウントします〜~
def get_counts(sequence):
counts = {}
for x in sequence:
if x in counts:
counts[x] += 1
else:
counts[x] = 1
return counts
counts = get_counts(time_zones)
print counts['America/New_York']