Item Type |
Article |
ID |
|
Preview |
Image |
|
Caption |
|
|
Full text |
KO11003001-20230304-0025.pdf
Type |
:application/pdf |
Download
|
Size |
:987.6 KB
|
Last updated |
:Feb 20, 2024 |
Downloads |
: 32 |
Total downloads since Feb 20, 2024 : 32
|
|
Release Date |
|
Title |
Title |
Prototypical Contrastive Transfer Learning for multimodal language understanding
|
Kana |
|
Romanization |
|
|
Other Title |
|
Creator |
Name |
小槻, 誠太郎
|
Kana |
オツキ, セイタロウ
|
Romanization |
Otsuki, Seitaro
|
Affiliation |
慶應義塾大学理工学部情報工学科
|
Affiliation (Translated) |
Department of Information and Computer Science, Keio University
|
Role |
|
Link |
|
|
Edition |
|
Place |
|
Publisher |
Name |
慶應義塾大学AI・高度プログラミングコンソーシアム
|
Kana |
ケイオウ ギジュク ダイガク AI・コウド プログラミング コンソーシアム
|
Romanization |
Keiō gijuku daigaku AI kōdo puroguramingu konsōshiamu
|
|
Date |
Issued (from:yyyy) |
2023
|
Issued (to:yyyy) |
|
Created (yyyy-mm-dd) |
|
Updated (yyyy-mm-dd) |
|
Captured (yyyy-mm-dd) |
|
|
Physical description |
|
Source Title |
Name |
AICカンファレンス予稿集
|
Name (Translated) |
|
Volume |
|
Issue |
|
Year |
2023
|
Month |
|
Start page |
25
|
End page |
25
|
|
ISSN |
|
ISBN |
|
DOI |
|
URI |
|
JaLCDOI |
|
NII Article ID |
|
Ichushi ID |
|
Other ID |
|
Doctoral dissertation |
Dissertation Number |
|
Date of granted |
|
Degree name |
|
Degree grantor |
|
|
Abstract |
We focus on the task of identifying target objects in domestic environments according to free-form natural language instructions. In this work, we propose a novel transfer learning approach for multimodal language understanding, Prototypical Contrastive Transfer Learning (PCTL) which uses a new contrastive loss, Dual ProtoNCE. We introduce PCTL to the target task. To validate PCTL, we built new real-world and simulation datasets. Our experiment demonstrated that PCTL outperformed existing methods. Specifically, PCTL achieved an accuracy of 78.1%, while simple fine-tuning achieved an accuracy of 73.4%.
|
|
Table of contents |
|
Keyword |
Multimodal Language Understanding
|
Prototypical Contrastive Transfer Learning
|
|
NDC |
|
Note |
会議名 : AICカンファレンス2023
開催地 : 慶應義塾大学日吉キャンパス
日時 : 2023年3月4日
第2章ポスター発表要旨
ポスター要旨-2
|
|
Language |
|
Type of resource |
|
Genre |
|
Text version |
|
Related DOI |
|
Access conditions |
|
Last modified date |
|
Creation date |
|
Registerd by |
|
History |
|
Index |
|
Related to |
|