-
Notifications
You must be signed in to change notification settings - Fork 40
New issue
Have a question about this project? # for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “#”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? # to your account
[20210404] Weekly Arxiv 만담 #4
Comments
Efficient Linear Transformers with Kernel Methods: Rethinking Attention with Performers: Random Feature Attention: ICLR 2021에서 Kernel method을 사용해 self-attention의 O(N^2)을 O(N)으로 바꾸고자 하는 논문 2개가 oral session과 spotlight paper로 선정되었습니다. 구글과 딥마인드에서 나온 연구인데 softmax를 직접 연산하는 대신 kernel(SVM에서의 kernel과 동일한 kernel입니다)을 통해서 attention을 연산합니다. Transformer의 가장 큰 문제점 중 하나인 quadratic growth를 해결할 수 있는 방법론으로 앞으로 많은 발전을 이룰 것을 예상합니다. |
Unsupervised Hyperbolic Representation Learning via Message Passing Auto-Encoders paper ; https://arxiv.org/pdf/2103.16046.pdf 기존 Euclidean space 에서 진행하였던 embedding 과 본 논문 저자가 주장하는 hyperbolic 에서 진행한 embedding 방법론을 비교하고자 link prediction , node clustering task 를 통해 우수함을 보임. Question; Table 2에서 보인 link prediction performance 에서 Pubmed (bio) dataset 은 오히려 comparison 중 하나인 DBGAN이 더 우수하였습니다. 기존 분자 구조들은 3차원이라 hyperbolic space에서 좀 더 좋은 performance가 나올것이라 기대되었으나 그렇지않았는데 DBGAN paper을 읽어보며 왜 그러한 결과가 나왔는가에 탐구해보면 흥미로운 인사이트를 얻을 수 있지 않을까 기대가 됩니다. 혹 인사이트를 얻게 된다면 이 방에 공유토록 하겠습니다. :) |
|
For next week
|
Steven Boyd 교수님께서 Minimum DIstortion Embedding (MDE)라는 논문 및 라이브러리를 발표하셨습니다. 논문: https://arxiv.org/abs/2103.02559 |
AI News
ArXiv
EfficientNetV2: Smaller Models and Faster Training
Putting NeRF on a Diet: Semantically Consistent Few-Shot View Synthesis
A Survey on Natural Language Video Localization
Using Python for Model Inference in Deep Learning
Mesh Graphormer
CUPID: Adaptive Curation of Pre-training Data for Video-and-Language Representation Learning
Many-to-English Machine Translation Tools, Data, and Pretrained Models
Unsupervised Sound Localization via Iterative Contrastive Learning
Jigsaw Clustering for Unsupervised Visual Representation Learning
FeTaQA: Free-form Table Question Answering
Anchor Pruning for Object Detection
Enriched Music Representations with Multiple Cross-modal Contrastive Learning
SCALoss: Side and Corner Aligned Loss for Bounding Box Regression
Improving Calibration for Long-Tailed Recognition
In&Out : Diverse Image Outpainting via GAN Inversion
Is Label Smoothing Truly Incompatible with Knowledge Distillation: An Empirical Study
ReMix: Towards Image-to-Image Translation with Limited Data
VITON-HD: High-Resolution Virtual Try-On via Misalignment-Aware Normalization
The User behind the Abuse: A Position on Ethics and Explainability
사심을 담은 논문 2개
The text was updated successfully, but these errors were encountered: