Wonho Jhe

Joon-Hyuk Ko, Hankyul Koh, Nojun Park et al.

Neural Ordinary Differential Equations (NeuralODEs) present an attractive way to extract dynamical laws from time series data, as they bridge neural networks with the differential equation-based modeling paradigm of the physical sciences. However, these models often display long training times and suboptimal results, especially for longer duration data. While a common strategy in the literature imposes strong constraints to the NeuralODE architecture to inherently promote stable model dynamics, such methods are ill-suited for dynamics discovery as the unknown governing equation is not guaranteed to satisfy the assumed constraints. In this paper, we develop a new training method for NeuralODEs, based on synchronization and homotopy optimization, that does not require changes to the model architecture. We show that synchronizing the model dynamics and the training data tames the originally irregular loss landscape, which homotopy optimization can then leverage to enhance training. Through benchmark experiments, we demonstrate our method achieves competitive or better training loss while often requiring less than half the number of training epochs compared to other model-agnostic techniques. Furthermore, models trained with our method display better extrapolation capabilities, highlighting the effectiveness of our method.

Hankyul Koh, Joon-hyuk Ko, Wonho Jhe

We propose the Moderate Adaptive Linear Unit (MoLU), a novel activation function for deep neural networks, defined analytically as: f(x)=x \times (1+tanh(x))/2. MoLU combines mathematical elegance with empirical effectiveness, exhibiting superior performance in terms of prediction accuracy, convergence speed, and computational efficiency. Due to its C-infinity smoothness, i.e. infinite differentiability and analyticity, MoLU is expected to mitigate issues such as vanishing or exploding gradients, making it suitable for a broad range of architectures and applications, including large language models (LLMs), Neural Ordinary Differential Equations (Neural ODEs), Physics-Informed Neural Networks (PINNs), and Convolutional Neural Networks (CNNs). Empirical evaluations show that MoLU consistently achieves faster convergence and improved final accuracy relative to widely used activation functions such as GeLU, SiLU, and Mish. These properties position MoLU as a promising and robust candidate for general-purpose activation across diverse deep learning paradigms.

QHwan Kim, Joon-Hyuk Ko, Sunghoon Kim et al.

The characterization of drug-protein interactions is crucial in the high-throughput screening for drug discovery. The deep learning-based approaches have attracted attention because they can predict drug-protein interactions without trial-and-error by humans. However, because data labeling requires significant resources, the available protein data size is relatively small, which consequently decreases model performance. Here we propose two methods to construct a deep learning framework that exhibits superior performance with a small labeled dataset. At first, we use transfer learning in encoding protein sequences with a pretrained model, which trains general sequence representations in an unsupervised manner. Second, we use a Bayesian neural network to make a robust model by estimating the data uncertainty. As a result, our model performs better than the previous baselines for predicting drug-protein interactions. We also show that the quantified uncertainty from the Bayesian inference is related to the confidence and can be used for screening DPI data points.