Time series comparison, anomaly detection, classification (supervised learning), and clustering (unsupervised learning) based on dynamic and second-order dependence structures
Learning from data in both the time and frequency domains
Scalable methods for large collections of time series
Applications to continuous health monitoring using wearable and sensor data
II. Streaming Data and Online Statistical Learning
• Statistical inference for streaming data with continuous arrival of observations
• One-pass, memory-efficient online algorithms for variance and covariance matrix estimation
• Adaptive spectral variance estimation for large-scale Monte Carlo and MCMC simulation data
Pattern of hormone levels at different stages
Heart beat rates: differences in the frequency patterns
III. Machine Learning, AI, and Health Data Science (Asthma and the NIH All of Us Program)
The development and application of machine learning and artificial intelligence methods for large-scale health data, with a particular emphasis on asthma surveillance and health disparities in Hispanic and rural communities. Leveraging the NIH All of Us Researcher Workbench and other large-scale cohort datasets, I work with high-dimensional, heterogeneous, and longitudinal data to develop scalable, interpretable, and robust learning algorithms. This includes big data analytics, supervised and unsupervised learning, feature engineering, and AI-based risk stratification models for population-level inference and public health decision support.
Damage detection of a mechanical system
IV. Statistical Finance and Prediction
Financial time series with nonlinear dynamics and heavy-tailed behavior
New distance metrics for capturing nonlinear dependence in time series classification and clustering
Dynamic feature extraction and learning models for volatility modeling and prediction
Detection of structural changes and assessment of stationarity using statistical and machine learning approaches
V. Develop scalable machine learning and data science frameworks for spatiotemporal learning, forecasting, and representation of environmental systems, integrating heterogeneous data sources (cruises, Wave Gliders, in-situ sensors, satellites) across multiple resolutions. By combining physics-informed ML, data assimilation, and adaptive filtering, this work aims to model the four-dimensional dynamics of pH, Ω_arag, and related chemical variables, while discovering interpretable relationships among environmental drivers and quantifying uncertainty in the Gulf of Mexico.
VI. Longitudinal data analysis via semi-parametric partially linear models to study how CD4 cell numbers depends on different factors.
