Wonjae Lee
Wonjae Lee has a total of 8 co-authors. Here are the top co-authors by number of shared publications:
- Kangwon Seo
University of Missouri
2 shared publications - Cheng-Bang Chen
Affiliation Department of Industrial and Systems Engineering, University of Miami, Coral Gables, Florida, United States of America
1 shared publication - Hyeong Suk Na
University of Missouri
1 shared publication - Jae-Hyuk Choi
Division of Marine System Engineering, Korea Maritime and Ocean University, Busan, South Korea
1 shared publication Kyung-Sun Lim
1 shared publication- Long Zhao
Electrical Engineering and Computer Science Department, South Dakota School of Mines and Technology, Rapid City, SD, USA
1 shared publication
Latest Publications
- Open Access
- 14 September 2026
- Published by Elsevier in SSRN Electronic Journal
Oracle-Free Active Learning: A Confidence-Gated Generative Augmentation Framework for Imbalanced Classification
Every acquisition function in common use rests on an assumption class-conditional generation violates: that a selected sample's label is unknown and must be bought from an oracle. That assumption is what makes uncertainty the quantity to maximise. When candidates are drawn from a generator that was told which class to synthesise, the label is free, and the criterion inverts --- an uncertain candidate is no longer information awaiting purchase but a sample whose assigned class is probably wrong, that is, label noise. This paper takes that inversion as its premise and develops oracle-free acquisition for generative augmentation under severe class imbalance: the candidates worth keeping are those a classifier confidently assigns to the class they were conditioned on, and are mutually diverse. The rule puts the classifier inside the generative loop rather than after it, realised in four coupled stages --- an autoencoder that initialises the generator and supplies a class-conditional latent prior, a gradient-penalised adversarial backbone, interpolation-seeded candidates filtered by the oracle-free rule, and co-training of critic, generator and classifier anchored to a frozen reference that stops the generator chasing an imperfect classifier. An ablation attributes performance to each stage. Against five established generative augmenters on three public image datasets under a shared protocol, the method ranks first on all four metrics on two datasets and leads six of the nine cells of an imbalance-ratio sweep. The results further show that realism and utility are separate objectives: the augmenter with the best Fr\'echet Inception Distance returns the worst macro-$F2$.
- 14 September 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessOracle-Free Active Learning: A Confidence-Gated Generative Augmentation Framework for Imbalanced Classification
Every acquisition function in common use rests on an assumption class-conditional generation violates: that a selected sample's label is unknown and must be bought from an oracle. That assumption is what makes uncertainty the quantity to maximise. When candidates are drawn from a generator that was told which class to synthesise, the label is free, and the criterion inverts --- an uncertain candidate is no longer information awaiting purchase but a sample whose assigned class is probably wrong, that is, label noise. This paper takes that inversion as its premise and develops oracle-free acquisition for generative augmentation under severe class imbalance: the candidates worth keeping are those a classifier confidently assigns to the class they were conditioned on, and are mutually diverse. The rule puts the classifier inside the generative loop rather than after it, realised in four coupled stages --- an autoencoder that initialises the generator and supplies a class-conditional latent prior, a gradient-penalised adversarial backbone, interpolation-seeded candidates filtered by the oracle-free rule, and co-training of critic, generator and classifier anchored to a frozen reference that stops the generator chasing an imperfect classifier. An ablation attributes performance to each stage. Against five established generative augmenters on three public image datasets under a shared protocol, the method ranks first on all four metrics on two datasets and leads six of the nine cells of an imbalance-ratio sweep. The results further show that realism and utility are separate objectives: the augmenter with the best Fr\'echet Inception Distance returns the worst macro-$F2$.
- 12 May 2025
- Published by Elsevier in Sustainable Energy, Grids and Networks
- Vol. 43
Clustering electricity consumption patterns using functional data analysis
- 12 May 2025
- Published by Elsevier in Sustainable Energy, Grids and Networks
- Vol. 43
Clustering electricity consumption patterns using functional data analysis
- Open Access
- 26 April 2022
- Published by Elsevier in Big Data Research
- Vol. 28
Downsampling for Binary Classification with a Highly Imbalanced Dataset Using Active Learning
- 26 April 2022
- Published by Elsevier in Big Data Research
- Vol. 28
Open AccessDownsampling for Binary Classification with a Highly Imbalanced Dataset Using Active Learning