Generative Artificial Intelligence in Biomedicine
Generative AI produces data with patterns and structures similar to the input data. We leverage these models for creating medical documentation. Additionally, we develop Large Language Models for biomarker discovery using various genomic data types in text and signals.
Graph-based Machine Learning for Multi-modal Data
We develop graph-based deep learning models to discover novel biomarkers using various genomic resources. Additionally, we construct disease risk prediction models based on multi-modal data from patients, utilizing a graph-based representation.
Machine Learning for Tabular Data
A significant amount of data in manufacturing, finance, and medicine exists in tabular form. While there are numerous advanced deep learning models for image and text data, these methods are not necessarily superior to traditional machine learning approaches for tabular data. We focus on developing strategies to enhance the performance of machine learning models for tabular data, including supervised learning and data imputation. These models are integrated as reliable and interpretable AI systems.
Data Integration and Visualization
Optimizing all processing steps and linking datasets for a certain type of analysis is complicated and provides a significant impact to scientific community. We develop software pipelines for integrating various massive datasets. Visualization is also important to help for users to make a decision. We design efficient ways for showing important features and their correlations learned from data.