PTT數位生活區 / Prob_Solve (計算數學 Problem Solving)

[問題] A3C Actor Gradient

看板Prob_Solve (計算數學 Problem Solving)作者 (長鷹寶寶實驗室)時間7年前 (2017/10/08 10:31), 編輯推噓0(0推 0噓 5→)
留言5則, 1人參與, 7年前最新討論串1/1
Working on A3C deep reinforcement learning. Since I am too lazy to modify the last layer of my NN to softmax, I use a softmax filter to let the linear layer directly target the softmax output. The algorithm works in my test cases for now. But it might go wrong when the reward is on a different scale. Can anyone help me to check if my implementation is correct? https://goo.gl/FV8sFu -- ※ 發信站: 批踢踢實業坊(ptt.cc), 來自: 114.35.245.133 ※ 文章網址: https://www.ptt.cc/bbs/Prob_Solve/M.1507429910.A.70E.html
→
11/05 22:07, 7年前 , 1F
It turns out that the current test case will
11/05 22:07, 1F
→
11/05 22:08, 7年前 , 2F
make correct result target to 1 after softmax
11/05 22:08, 2F
→
11/05 22:08, 7年前 , 3F
and the wrong result to 0
11/05 22:08, 3F
→
11/05 22:09, 7年前 , 4F
That's why the reward will work in its current
11/05 22:09, 4F
→
11/05 22:09, 7年前 , 5F
scale
11/05 22:09, 5F
‣ 更多 longlyeagle 的文章
文章代碼(AID): #1PsOuMSE (Prob_Solve)
Prob_Solve 近期熱門文章
更多 近期熱門文章 >>
PTT數位生活區 即時熱門文章
更多 即時熱門文章 >>
‣ 更多 longlyeagle 的文章
文章代碼(AID): #1PsOuMSE (Prob_Solve)