본문 바로가기

Papers

(12)
[리뷰] Structured Prediction as Translation between Augmented Natural Languages (ICLR 2021) by Paolini et al. Paper Link : https://openreview.net/pdf?id=US-TP-xnXI Code Link : https://github.com/amazon-research/tanl Contents 1. Structured Prediction 2. Structured Prediction Tasks 3. Introduction 4. Proposed Method 5. Experiments 6. Discussion & Summary Abbreivation & Keywords SP : Structured Prediction TANL : Translation between Augmented Natural Language PLM : Pre-trained Language Mod..
[리뷰] In Defense of Pseudo-Labeling: An Uncertainty-Aware Pseudo-label Selection Framework for Semi-Supervised Learning (ICLR 2021) by Rizve et al. Paper Link: https://arxiv.org/pdf/2101.06329.pdf Code Link : https://github.com/nayeemrizve/ups Contents 1. Introduction 2. Related Works 3. Proposed Method 4. Experiments 5. Discussion 6. Summary Abbreviation & Keyword SSL : Semi-supervied Learning PL : Pseudo-labeling CR : Consistency-regularization UPS : Uncertainty-aware pseudo-label selection framework Introduction Deep Lear..
[리뷰] Interpreting Graph Neural Networks for NLP With Differentiable Edge Masking (ICLR 2021) by Schlichtkrull et al. Paper Link : https://openreview.net/pdf?id=WznmQa42ZAx Contents 1. Introduction 2. Related Works 3. Methods 4. Synthetic Experiment 5. Question Answering 6. Semantic Role Labeling 7. Summary & Conclusion Abstract GNNs are famous apporach for integrating structural inductive biases into NLP models There is little work on interpreting for understanding which parts of the gr..
[리뷰] Distilling the Knowledge in a Neural Network (NIPS Workshop 2014) by Hinton et al. Paper Link : https://arxiv.org/pdf/1503.02531.pdf Contents 1. Introduction 2. Distillation 3. Experiments (MNIST, Speech Recognition) 4. Training Ensembles on Very Big Datasets (JFT) 5. Soft Targets as Regularizers 6. Summary Introduction 다수의 사용자가 사용하는 서비스를 배포하고자 할 때, latency와 computational resources는 중요한 문제입니다. 예를 들어, ensemble 모델은 모바일 프로그램으로 배포하기에 부적합할 수 있습니다. 이러한 경우 도입해볼 수 있는 ..

반응형