Skip to content

Latest commit

 

History

History
3 lines (2 loc) · 1.06 KB

diverse-beam-search.md

File metadata and controls

3 lines (2 loc) · 1.06 KB

Diverse Beam Search: Decoding Diverse Solutions from Neural Sequence Models

TLDR; The authors propose a new Diverse Beam Search (DBS) decoding procedure that produces more diverse responses than standard Beam Search (BS). The authors divide the beam of size B into G groups of size B/G. At each step they perform beam search for each group with an added similarity penalty (with scaling factor lambda) that encourages groups to be pick different outputs. This procedure is done greedily, i.e. group 1 does regular BS, group 2 is conditioned on group 1, group 3 is conditioned on group 1 and 2, and so on. Similarity functions include Hamming distance, Cumulative Diversity, n-gram diversity and neural embedding diversity. Hamming Distance tends to perform best. The authors evaluate their model on Image Captioning (COCO, PASCAL-50S), Machine Translation (europarl) and Visual Question Generation. For Image Captioning the authors perform a human evaluation (1000 examples on Mechanical Turk) and find that DBS is preferred over BS 60% of the time.