Seungyub Han

I am a PhD student at the Communications and Machine Learning Laboratory, part of the Department of Electrical and Computer Engineering at Seoul National University, where I work on Reinforcement Learning, Robotics, and Deep Learning. My PhD advisor is Jungwoo Lee.

I have a BS in EE from Seoul National University.

E-mail: seungyubhan@snu.ac.kr

GitHub  /  Google Scholar  /  LinkedIn  / 

Publications

I'm interested in reinforcment learning, robot learning, optimization, and representation learning.

Pitfall of Optimism: Distributional Reinforcement Learning by Randomizing Risk Criterion

We provide a perturbed distributional Bellman optimality operator by distorting the risk measure and prove the convergence and optimality of the proposed method with the weaker contraction property.

SPQR: Controlling Q-ensemble Independence with Spiked Random Model for Reinforcement Learning

By introducing a novel regularization loss for Q-ensemble independence based on random matrix theory, we propose spiked Wishart Q-ensemble independence regularization (SPQR) for reinforcement learning.

On the Convergence of Continual Learning with Adaptive Methods

In this paper, we provide a convergence analysis of memory-based continual learning with stochastic gradient descent and empirical evidence that training current tasks causes the cumulative degradation of previous tasks.

Learning to Learn Unlearned Feature for Brain Tumor Segmentation

One of the difficulties in medical image segmentation is the lack of datasets with proper annotations. To alleviate this problem, we propose active meta-tune to achieve balanced parameters for both glioma and brain metastasis domains within a few steps.

Preprints

Generative Adversarial Trainer: Defense to Adversarial Perturbations with GAN

We propose a novel technique to make neural network robust to adversarial examples using a generative adversarial network. We alternately train both classifier and generator networks. The generator network generates an adversarial perturbation that can easily fool the classifier network by using a gradient of each image.

Other Activities

Design and source code from Jon Barron's website