AdaBlock: SGD with Practical Block Diagonal Matrix Adaptation for Deep Learning

AISTATS 2022 AdaBlock: SGD with Practical Block Diagonal Matrix Adaptation for Deep Learning Jihun Yun, Aurelie Lozano, Eunho Yang  We introduce AdaBlock, a class of adaptive gradient methods that extends popular approaches such as Adam by adopting the simple and natural idea of ...

A Simple Framework for Robust Out-of-Distribution Detection

IEEE 2022 A Simple Framework for Robust Out-of-Distribution Detection  Youngbum Hur, Eunho Yang, Sung Ju Hwang  Out-of-distribution (OOD) detection, i.e., identifying whether a given test sample is drawn from outside the training distribution, is essential for a deep cl...

Unbiased Classification through Bias-Contrastive and Bias-Balanced Learning

NeurIPS 2021 Unbiased Classification through Bias-Contrastive and Bias-Balanced Learning Youngkyu Hong, Eunho Yang  Datasets for training machine learning models tend to be biased unless the data is collected with complete care. In such a biased dataset, models are susceptible to...

Task-Adaptive Neural Network Retrieval with Meta-Contrastive Learning

NeurIPS 2021 Task-Adaptive Neural Network Retrieval with Meta-Contrastive Learning Wonyong Jeong, Hayeon Lee,Gun Park, Eunyoung Hyung, Jinheon Baek andSung Ju Hwang  Most conventional Neural Architecture Search (NAS) approaches are limited in that they only generate architectures...

RetCL: A Selection-based Approach for Retrosynthesis via Contrastive Learning

IJCAI 2021 RetCL: A Selection-based Approach for Retrosynthesis via Contrastive Learning Hankook Lee, Sungsoo Ahn, Seung-Woo Seo, You Young Song, Eunho Yang, Sung Ju Hwang and Jinwoo Shin  Retrosynthesis, of which the goal is to find a set of reactants for synthesizing a target p...

Rapid Neural Architecture Search by Learning to Generate Graphs from Datasets

ICLR 2021 Rapid Neural Architecture Search by Learning to Generate Graphs from Datasets Hayeon Lee, Eunyoung Hyung and Sung Ju Hwang  Despite the success of recent Neural Architecture Search (NAS) methods on various tasks which have shown to output networks that largely outperfo...

Mutually-Constrained Monotonic Multihead Attention for Online ASR

ICASSP 2021 Mutually-Constrained Monotonic Multihead Attention for Online ASR  Jaeyun Song, Hajin Shim, Eunho Yang   Despite the feature of real-time decoding, Monotonic Multihead Attention (MMA) shows comparable performance to the state-of-the-art offline methods ...

Multi-domain Knowledge Distillation via Uncertainty-Matching for End-to-End ASR Models

INTERSPEECH 2021 Multi-domain Knowledge Distillation via Uncertainty-Matching for End-to-End ASR Models Ho-Gyeong Kim, Min-Joong Lee, Hoshik Lee, Tae Gyoon Kang, Jihyun Lee, Eunho Yang and Sung Ju Hwang  Knowledge Distillation basically matches predictive distributions of st...

Mini-Batch Consistent Slot Set Encoder for Scalable Set Encoding

NeurIPS 2021 Mini-Batch Consistent Slot Set Encoder for Scalable Set Encoding Bruno Andreis, Jeffrey Willette, Juho Lee and Sung Ju Hwang  Most existing set encoding algorithms operate under the assumption that all the elements of the set are accessible during training and infere...

Meta-StyleSpeech : Multi-Speaker Adaptive Text-to-Speech Generation

ICML 2021 Meta-StyleSpeech : Multi-Speaker Adaptive Text-to-Speech Generation Dongchan Min, Dong Bok Lee, Eunho Yang and Sung Ju Hwang  With rapid progress in neural text-to-speech (TTS) models, personalized speech generation is now in high demand for many applications. For pract...