Online Hyperparameter Meta-Learning with Hypergradient Distillation

ICLR 2022 Online Hyperparameter Meta-Learning with Hypergradient Distillation Hae Beom Lee, Hayeon Lee, Jaewoong Shin, Eunho Yang, Timothy Hospedales, Sung Ju Hwang    Many gradient-based meta-learning methods assume a set of parameters that do not participate in inner-optim...

Online Coreset Selection for Rehearsal-based Continual Learning

ICLR 2022 Online Coreset Selection for Rehearsal-based Continual Learning Jaehong Yoon, Divyam Madaan, Eunho Yang, Sung Ju Hwang  A dataset is a shred of crucial evidence to describe a task. However, each data point in the dataset does not have the same potential, as some of...

On Divergence Measures for Bayesian Pseudocoresets

NeurIPS 2022 On Divergence Measures for Bayesian Pseudocoresets  Balhae Kim, Jungwon Choi, Seanie Lee, Yoonho Lee, Jung-Woo Ha, Juho Lee   A Bayesian pseudocoreset is a small synthetic dataset for which the posterior over parameters approximates that of the origina...

MPViT: Multi-Path Vision Transformer for Dense Prediction

CVPR 2022 MPViT: Multi-Path Vision Transformer for Dense Prediction Youngwan Lee, Jonghee Kim, Jeff Willette, Sung Ju Hwang    Dense computer vision tasks such as object detection and segmentation require effective multi-scale feature representation for detecting or classify...

Model-augmented Prioritized Experience Replay

ICLR 2022 Model-augmented Prioritized Experience Replay Youngmin Oh, Jinwoo Shin, Eunho Yang, Sung Ju Hwang    Experience replay is an essential component in off-policy model-free reinforcement learning (MfRL). Due to its effectiveness, various methods for calculating priori...

Meta Learning Low Rank Covariance Factors for Energy Based Deterministic Uncertainty

ICLR 2022 Meta Learning Low Rank Covariance Factors for Energy Based Deterministic Uncertainty Jeffrey Ryan Willette, Hae Beom Lee, Juho Lee, Sung Ju Hwang  Numerous recent works utilize bi-Lipschitz regularization of neural network layers to preserve relative distances between d...

Learning the Compositional Domains for Generalized Zero-shot Learning

Computer Vision and Image Understanding vol.221 Learning the Compositional Spaces for Generalized Zero-shot Learning Hanze Dong, Yanwei Fu, Sung Ju Hwang, Leonid Sigal, Xiangyang Xue  This paper studies the problem of Generalized Zero-shot Learning (G-ZSL), whose goal is to class...

KALA: Knowledge-Augmented Language Model Adaptation

NAACL 2022 KALA: Knowledge-Augmented Language Model Adaptation Minki Kang, Jinheon Baek, Sung Ju Hwang   Pre-trained language models (PLMs) have achieved remarkable success on various natural language understanding tasks. Simple fine-tuning of PLMs, on the other hand, might be s...

Improving Ensemble Distillation With Weight Averaging and Diversifying Perturbation

ICML 2022 Improving Ensemble Distillation With Weight Averaging and Diversifying Perturbation  Giung Nam, Hyungi Lee, Byeongho Heo, Juho Lee   Ensembles of deep neural networks have demonstrated superior performance, but their heavy computational cost hinders applying t...

GraphENS: Neighbor-Aware Ego Network Synthesis for Class-Imbalanced Node Classification

ICLR 2022 GraphENS: Neighbor-Aware Ego Network Synthesis for Class-Imbalanced Node Classification Joonhyung Park*, Jaeyun Song*, Eunho Yang (*: equal contribution)  In many real-world node classification scenarios, nodes are highly class-imbalanced, where graph neural networks (G...