My research began with pruning and transfer learning for convolutional and artificial neural networks, which gave me a stronger practical understanding of how neural networks learn, adapt, and can be made more efficient.

My recent work brings neural-network and optimization methods to financial and tabular domains, with an emphasis on improving performance, efficiency, and interpretability.

Selected Publications

Multi-Objective Evolution of Patch Tokens for Tabular Transformers

Han Joon Byun, Joon Huh, Jieun Yook, and Byung-ro Moon

Proceedings of the Genetic and Evolutionary Computation Conference (GECCO '26)

2026

[DOI]

Evolutionary Feature Construction for Tabular Transformers via Gradient-Enhanced Attention Rollout

Han Joon Byun, Jieun Yook, Hanbin Koo, and Byung-ro Moon

Proceedings of the Genetic and Evolutionary Computation Conference Companion (GECCO '26 Companion)

2026

[DOI]

Label-Noise Decay Confounds Loss-Based Drift Attribution: A False-Positive Channel and a Noise-Aware Diagnostic

Joon Huh and Han Joon Byun

Transactions on Machine Learning Research (TMLR)

2026

[Paper]

Evolutionary Pruning of Deep Convolutional Networks by a Memetic GA with Sped-Up Local Optimization and GLCM Energy Z-Score

Ha Hyun Cho, Han Joon Byun, Min Kee Kim, Joon Huh, and Byung-Ro Moon

Proceedings of the Genetic and Evolutionary Computation Companion (GECCO '23 Companion)

2023

[DOI]