References & Citations
Computer Science > Computer Vision and Pattern Recognition
Title: A Better Variant of Self-Critical Sequence Training
(Submitted on 22 Mar 2020 (v1), last revised 10 May 2020 (this version, v2))
Abstract: In this work, we present a simple yet better variant of Self-Critical Sequence Training. We make a simple change in the choice of baseline function in REINFORCE algorithm. The new baseline can bring better performance with no extra cost, compared to the greedy decoding baseline.
Submission history
From: Ruotian Luo [view email][v1] Sun, 22 Mar 2020 19:04:25 GMT (72kb,D)
[v2] Sun, 10 May 2020 21:59:36 GMT (72kb,D)
Link back to: arXiv, form interface, contact.