Yao Zheng
University of Connecticut, CT·Mathematics & Statistics·Early-career
Time series, data recorded over time such as sensor readings or economic indicators, are now often high-dimensional, shifting and messy. Flexible machine learning models can forecast such data well, but they are hard to interpret and rarely say how uncertain their answers are.
- Fields
Mathematics & Statistics
Computer Science
- Topics
time series analysis
statistics
statistical inference
- Grants
- 1 known grant, through Aug 2030
What they work on
- Develop recurrent neural network (RNN) models for high-dimensional time series that stay interpretable through Granger causality, a test of whether one series predicts another, to balance accuracy and interpretability.
- Create an online, distribution-free procedure for adaptive model selection in nonstationary time series, controlling uncertainty through conformal miscoverage rate calibration.
- Design methods to combine time series of different granularities and to impute missing data under general patterns, using factor models and transfer learning.
More about the research
This project develops statistical models, theory and inference methods for modern time series. It addresses many variables at once, nonlinear and nonstationary behavior (patterns that change over time), different time resolutions, and missing or uneven-quality data.
Why it matters
Methods that are both accurate and interpretable could help analysts trust predictions from high-dimensional data. Real-time model selection with reliable uncertainty could support decisions as conditions change. Tools for merging data of different quality and resolution could make better use of time series collected from diverse sources.
Known grants
From NIH and NSF research grants. Labs often have other funding we can't see.
Closest labs
Benjamin Boniece
Drexel University, PA·Mathematics & Statistics
Develop methods to detect shocks that affect many series in high-dimensional, high-frequency data, possibly asynchronously, using time-localized principal component analysis in factor-structured models.
Ting Zhang
University of Georgia, GA·Mathematics & Statistics
Develop change-point tests that stay powerful when the post-change signal varies unpredictably, using a causal representation framework and Gaussian approximation to set statistically sound alarm thresholds.
Chao Zhang
Georgia Institute of Technology, GA·Computer Science·Early-career
Develop deep predictive models for spatial networks that capture how they change over space and time while estimating the uncertainty of their own forecasts.
The school
University of Connecticut: 31 labs on PI Pages, $434M a year on research, 360 research doctorates a year.
Reach out
A start for your email. Fill in the brackets.