Xiongye Xiao

Assistant Professor

Electrical Engineering and Computer Science
University of Tennessee, Knoxville

Foundational AI cluster AI Tennessee NSF STC COMPASS

AI and complex systems

I develop mathematical and AI methods for the quantitative study of structure and dynamics across scales. I ask what we can predict about each, and how they shape function. AI is both a method I develop and a complex system I study. I am extending this work toward inverse design.

I joined UTK in Fall 2025 and lead the DUAL research group. I studied at the University of Southern California (Ph.D., 2019–2025) and Zhejiang University (bachelor's, 2015–2019).

Email Scholar Research

Research at DUAL

Our group

Forward ModelingandInverse Design:Quantifying structure and dynamics across scales, and guiding design.AI for ScienceandScience for AI:Understanding the world with AI, and understanding AI itself.

Selected publications

All publications

Neuron-based Multifractal Analysis of Neuron Interaction Dynamics in Large Models

Xiongye Xiao, Heng Ping, ..., Paul Bogdan

ICLR 2025

NeuroMFA characterizes the multifractal organization of AI weight networks across training checkpoints.

NeuroMFA representation: from a neural network to a weighted neuron interaction network

Paper Code

MIHC: Multi-view Interpretable Hypergraph Neural Networks with Information Bottleneck for Chip Congestion Prediction

Zeyue Zhang, Heng Ping, ..., Xiongye Xiao

NeurIPS 2025

MIHC combines multi-view hypergraphs and an information bottleneck for interpretable congestion prediction.

Complete MIHC framework: circuit netlists and layouts, bottleneck subgraphs, multi-view hypergraph learning and congestion prediction

Paper

Spiking dynamics of individual neurons reflect changes in the structure and function of neuronal networks

Ruochen Yang, Heng Ping, Xiongye Xiao, Roozbeh Kiani, and Paul Bogdan

Nature Communications 2025

Multifractal spiking statistics connect neuronal dynamics to network structure and function.

Co-corresponding authors: Xiongye Xiao, Roozbeh Kiani, and Paul Bogdan.

Original Figure 1 connecting network topology with neuronal spike-train dynamics

Paper

Neuro-Inspired Information-Theoretic Hierarchical Perception for Multimodal Learning

Xiongye Xiao, Gengshuo Liu, ..., Paul Bogdan

ICLR 2024

ITHP combines a neuro-inspired processing hierarchy with information bottlenecks for multimodal learning.

ITHP: primary modality X0 passes through two encoders to a predictor; detectors for X1 and X2 guide training through information constraints

Paper Code

Coupled Multiwavelet Operator Learning for Coupled Differential Equations

Xiongye Xiao, Defu Cao, ..., Paul Bogdan

ICLR 2023

CMWNO learns coupled solution operators through interacting kernels in multiwavelet space.

Coupled multiwavelet neural operator decomposition and reconstruction architecture

Paper Code

Deciphering the Generating Rules and Functionalities of Complex Networks

Xiongye Xiao, Hanlong Chen, and Paul Bogdan

Scientific Reports 2021

NMFA uses node-centered box growth to characterize multiscale network structure and compare its complexity and heterogeneity.

Paper

Teaching & mentorship

I teach Artificial Intelligence (COSC 423/523) and Introduction to Machine Learning (COSC 325) at UTK.

I encourage students to turn broad interests into precise questions and testable hypotheses.

Selected recognition

2025
Order of Areté USC's highest graduate honor at commencement
2023
Best Poster Award (1st Prize) Fall Fourier Talks, University of Maryland
2022
Best Research Assistant Award University of Southern California
2022
Outstanding Research Assistant Award University of Southern California