Welcome to Chicago Human+AI Lab (CHAI)!
Our goal is to build the best AI for humans. We are most interested in the following applications (ordered randomly):
We are always looking for motivated postdocs, PhD students, and undergraduates who are interested in NLP, data science, computational social science, or machine learning! Please read this FAQ if you are interested.
Schedule and Syllabus for Human-Centered Machine learning.
This is the official repository for HypoGeniC (Hypothesis Generation in Context) and HypoRefine, which are automated, data-driven tools that leverage large language models to generate hypothesis fo…
Active Example Selection for In-Context Learning (EMNLP'22)
Introduces Iterated Function Networks as a graph-based generalization of iterated function systems, proving fundamental convergence properties and establishing their applications in fractal generation and optimization.
Extends classical iterated function systems to network structures by developing a mathematical framework that proves convergence properties and establishes bounds based on network topology.
Introduces α-smoothness property for utility functions in distributed pattern mining and proves it is necessary and sufficient for efficient convergence, providing the first complete theoretical characterization of when distributed pattern mining algorithms can be efficient.
Introduces omega-recursive sequence spaces as a new mathematical framework that connects recursive function theory with functional analysis, establishing completeness properties and hierarchical relationships.
Provides a complete classification of polynomial sequences generated by iterated rational functions over finite fields, proving they either become periodic or grow exponentially.
Introduces hybrid spectral operators (HSOs) that combine continuous and discrete spectral properties, proving their spectrum can be decomposed into continuous and discrete parts with specific lattice structure properties.
Establishes theoretical upper bounds on the probability of generating novel mathematical structures using entropy and topological invariants of the solution space.
Proves that spectral graph neural networks can approximate any arithmetic function with guaranteed precision and numerical stability bounds.
This organization has no public members. You must be a member to see who’s a part of this organization.
Loading…