论文阅读笔记
Araloak
这个作者很懒,什么都没留下…
展开
-
阅读笔记:Learning to Remember Rare Events
Learning to Remember Rare EventsContributionNMT难以记住在train set上的rare word,导致很多包括one-shot在内的场景表现不好,本文提出一个memory module可以用在NMT和CV中多种领域和模型上,作为主题模型的附加模块,提高表现Our module remembers training examples shown many thousands of steps in the past and it can succes原创 2020-12-15 14:29:41 · 430 阅读 · 1 评论 -
阅读笔记:ViLBERT: Pretraining Task-Agnostic Visiolinguistic Representations for Vision-and-Language Task
阅读笔记:ViLBERT: Pretraining Task-Agnostic Visiolinguistic Representations for Vision-and-Language TasksContribution提出 ViLBERT 模型(two streams model),由两个BERT结构分别对text和image进行学习,通过cross-attention进行信息交流,在两个预训练任务(proxy tasks)上进行预训练。最后在4个task上进行finetune:visua原创 2020-12-14 20:19:12 · 749 阅读 · 3 评论 -
阅读笔记:XGPT: Cross-modal Generative Pre-Training for Image Captioning
XGPT: Cross-modal Generative Pre-Training for Image CaptioningContribution现有大多数VL pre-trained models基本都是Transformer-Encoder结构的,不适用于Vision-and-language generation tasks,因为:On one hand, pre-trained models developed for understanding tasks only provide原创 2020-12-14 20:24:34 · 1289 阅读 · 1 评论 -
阅读笔记:VL-BERT: PRE-TRAINING OF G ENERICVISUAL-LINGUISTICR EPRESENTATIONS
阅读笔记:VL-BERT: PRE-TRAINING OF G ENERICVISUAL -LINGUISTICR EPRESENTATIONSContribution文章提出VL-BERT(single stream model),结合文本和图片进行end-to-end预训练,对一系列下游的图片—文本任务都有明显提升效果( image captioning、 visual question answering、 visual commonsense reasoning)从前做text与ima原创 2020-12-14 20:16:23 · 446 阅读 · 0 评论