Weijing Tang
Carnegie Mellon University, PA·Mathematics & Statistics
Modern AI systems learn from huge amounts of data, and many real-world datasets record relationships that are more complicated than a simple yes or no. Two statistical questions follow.
- Fields
Mathematics & Statistics
Computer Science
- Topics
data valuation
Shapley value
semi-values
- Grants
- 2 known grants, through Jun 2029
What they work on
- Develop priority-aware rules for valuing training data based on the Shapley value, a game-theory method for fairly splitting credit, that account for originality, provenance and individual risk within a unified axiomatic framework.
- Determine the statistical and computational limits of approximating Shapley values and semi-values for large, high-dimensional AI models, and design efficient estimation algorithms.
- Build a population-level theory of data value called Shapley density, a continuous version of data valuation, with convergence guarantees, faster computation and uncertainty quantification.
- Create statistical inference methods using nonparametric graphon models, flexible models of link probabilities, to test social theories of how signed networks with positive and negative ties form.
- Design fast community detection methods for signed networks under a hierarchical stochastic block model, and develop a latent-space framework for learning interpretable node embeddings.
More about the research
How much is each piece of training data worth to an AI model? And how should we analyze networks in which relationships can be friendly or hostile? Both projects build statistical theory and practical algorithms for these kinds of complex data.
Why it matters
As AI depends more on shared data, fair methods for valuing data could support transparent credit and payment for data contributors. They could also encourage sharing of high-quality data. Tools for networks with positive and negative ties could help researchers in political science, biochemistry, medicine, genetics, ecology and marketing analyze alliances, conflicts and competition in their systems.
Known grants
From NIH and NSF research grants. Labs often have other funding we can't see.
Closest labs
Yinqiu He
University of Wisconsin-Madison, WI·Mathematics & Statistics
Develop methods with theoretical guarantees for estimating low-dimensional network embeddings and quantifying their uncertainty under general network models with relaxed assumptions.
Marianna Pensky
University of Central Florida, FL·Mathematics & Statistics
Extend the multiplex Generalized Dot Product Graph model so that groups of network layers can have different underlying structures. This would allow layers to be clustered by condition.
Jonathan Stewart
Florida State University, FL·Mathematics & Statistics
Develop a new class of generalized latent position models for network data that drop the usual conditional independence assumption, so that dependence among connections, not just node positions, shapes the modeled network.
The school
Carnegie Mellon University: 28 labs on PI Pages, $505M a year on research, 346 research doctorates a year.
Reach out
A start for your email. Fill in the brackets.