【論文紹介】Deep Reinforcement Learning for Solving the Vehicle Routing Problem
1. Deep Reinforcement Learning for Solving the
Vehicle Routing Problem
Mohammadreza Nazari, Afshin Oroojlooy, Lawrence V. Snyder, Martin Taka ́cˇ
arXiv:1802.04240v1 [cs.AI] 12 Feb 2018
8. 𝑎" = 𝑎" 𝑥"
W
, ℎ" = 𝑠𝑜𝑓𝑡𝑚𝑎𝑥(𝑢")
𝑢"
W
= 𝑣`
a
tanh 𝑊`[𝑥"
W
; ℎ"]
𝑐" = h 𝑎"
W
𝑥"
W
i
jk#
P(𝑦"K# 𝑌", 𝑋" = 𝑠𝑜𝑓𝑡𝑚𝑎𝑥( 𝑢"
W
)
𝑢"
W
= 𝑣p
a
tanh 𝑊p[𝑥"
W
; 𝑐"]
埋め込み⼊⼒𝑥"
W
=(𝑠
W
, 𝑑"
W
)をiとし, ℎ"はデコーダの状態で整数ベクトル𝑎"の計算に使う
𝑢"
W
はtanℎをとった⽂脈ベクトルと埋め込み⼊⼒に重み付けしたもの
𝑃(𝑦"K# 𝑌", 𝑋" は𝑢"
W
のソフトマックス
Attention mechanismとは