About

Yaoyao Liu is an assistant professor in the School of Information Sciences and the Coordinated Science Laboratory at the University of Illinois Urbana-Champaign. He is also affiliated with the Siebel School of Computing and Data Science and the Department of Electrical & Computer Engineering. His research lies at the intersection of computer vision and machine learning, with a special focus on building intelligent visual systems that are continual and data-efficient. His research interests include continual learning, few-shot learning, semi-supervised learning, generative models, 3D geometry models, and medical imaging.

Research topics

• Physics
• Electronic engineering
• Engineering
• Atmospheric sciences
• Geology
• Materials science
• Electrical engineering
• Optoelectronics
• Composite material
• Metallurgy
• Meteorology
• Environmental science

Selected publications

• Spatial decay estimates for the coupled system of wave-plate type with thermal effect

  AIMS Mathematics · 2025-01-01 · 3 citations

  articleOpen accessSenior authorCorresponding

  <p>In this article, we investigate the spatial decay estimates for the biharmonic conduction equations within a coupled wave-plate system incorporating thermal effects in a two-dimensional cylindrical domain. Using the method of a second-order differential inequality, we can obtain the spatial decay estimates result for these equations. When the distance tends to infinity, the energy can decay exponentially. This result shows us that the Saint-Venant principle is also valid for the hyperbolic-parabolic coupled system.</p>

  PublisherDOI

• A GPU‐Accelerated Generative Adversarial Model for Causal Inference

  Concurrency and Computation Practice and Experience · 2025-08-08

  articleOpen accessSenior author

  ABSTRACT We develop a GPU‐accelerated machine learning generative adversarial model designed to facilitate causal inferences from observational data. Our model's theoretical framework is conceptualized in a manner that is amenable to being operable and scalable for high‐performance computing platforms. We leverage GPU acceleration to develop a parallel evolutionary algorithm to achieve large‐scale parallel computation of the model within a now widely accessible computing platform. This capability both enhances computational speedup and efficiency and also extends the use of the model to a broader range of substantive research domains while maintaining the underlying theoretical properties of the model.

  PublisherDOI

• Integrated Modeling Driven Evaluation of Opportunities for Climate‐Resilient Perennial Biomass Crop Plantings in Flood‐Prone Agricultural Landscapes

  Journal of Flood Risk Management · 2025-05-06

  articleOpen access

  ABSTRACT Adapting to future climate change in flood‐prone landscapes will require climate‐resilient agricultural systems. Planting perennial crops, like switchgrass and willow, along river corridors can mitigate future flooding while supporting bioenergy markets. We developed an integrated assessment linking climate, hydrologic, and inundation model results to assess future flood risk to river‐adjacent agricultural lands in the Mid‐Atlantic Region (MAR) and explore this opportunity. We produced ensemble streamflow projections for every MAR stream using a hydrologic model driven by a suite of downscaled and bias‐corrected Coupled Model Intercomparison Project Phase 6 climate projections. We then conducted high‐resolution inundation mapping based on projected flood frequencies for baseline and future periods. Results show that in the near‐term future, at least two‐thirds of the streams will experience 100‐year floods more severe than the baseline 200‐year floods. Riparian zones are projected to face a median rise of inundation by 9.5%–24.1%. Results show that there is an opportunity to mitigate flooding in over half of MAR's counties with the quantities of switchgrass and willow plantings anticipated for mature bioenergy markets, even under the most extreme (200‐year) flood events. Our integrated modeling framework can guide similar regions to evaluate opportunities for flood‐resilient agricultural systems under climate change.

  PublisherOA PDFDOI

• AvatarMirror: Rendering Restoration for Real-Time Immersive Presence System

  2025-03-08

  article

  Human-Centered immersive computing is catching on within VR and AR. We establish a real-time immersive presence system called AvatarMirror that enables arbitrary characters to look at themselves instantly in novel stereo views. First, stereo matching is implemented to acquire multi-view depth priors, based on which the human topology is reconstructed through raycasting from a novel perspective. Then, the color information from the referenced views is adaptively blended to perform the human rendering. Furthermore, we propose an RR-Net to conduct rendering restoration. AvatarMirror realizes real-time high-fidelity immersive presence with multi-threaded scheduling and CUDA acceleration.

  PublisherDOI

• A Transfer Learning Approach to Energy-Efficient Control of Small and Medium-Sized Commercial Buildings

  SSRN Electronic Journal · 2025-01-01

  preprintOpen access1st authorCorresponding
  PublisherDOI

• A novel nitrogen-induced graded cemented carbides with γ'-strengthened binder phase

  International Journal of Refractory Metals and Hard Materials · 2025-04-10 · 2 citations

  article
  PublisherDOI

• Non-Obese Hepatic Steatosis Severity Prediction: Machine Learning Model Development and Validation (Preprint)

  Journal of Medical Internet Research · 2025-08-17

  articleOpen access

  <sec> <title>BACKGROUND</title> Non-obese individuals account for 40% of global steatotic liver disease (SLD) cases, yet lack dedicated targeted screening tools. Current ultrasound-based methods exhibit low detection rates for mild steatosis, delaying intervention. </sec> <sec> <title>OBJECTIVE</title> This study aimed to develop and validate a non-invasive, multi-class machine learning model using the Ultrasound Attenuation Parameter to predict hepatic steatosis severity grades (none/mild/moderate-severe) in non-obese populations. It sought to address the critical diagnostic gap in early detection and risk stratification for this under-recognized group, thereby enabling timely intervention. </sec> <sec> <title>METHODS</title> A cohort of 215,145 participants enrolled from 2018 to 2024 was analyzed. UAP thresholds defined steatosis severity: &lt;244 dB/m (none), 244–269 (mild), &gt;269 (moderate to severe). Least Absolute Shrinkage and Selection Operator (LASSO) regression identified 14 predictors. Six ML models were trained (70% of the dataset) and validated (30%) using 10-fold cross-validation. Performance metrics included accuracy, Cohen’s kappa, area under the receiver operating characteristic curve (AUROC), F1-score, and SHapley Additive exPlanations (SHAP) analysis for interpretability. </sec> <sec> <title>RESULTS</title> The ML models were developed and validated using 150,602 participants in the training set and 64,543 in the test set, comprising non-SLD (n=92,944), mild SLD (n=54,121), and moderate-to-severe SLD (n=68,080). The Extreme Gradient Boosting (XGBoost) model demonstrated superior performance compared to other models. On the training set, it achieved a macro-average AUROC of 0.929, a macro-average precision-recall (PR) AUC of 0.878, and an accuracy of 0.788. On the test set, performance remained strong, with a macro-average AUROC of 0.908, a macro-average PR AUC of 0.842, and an accuracy of 0.759, surpassing other models. </sec> <sec> <title>CONCLUSIONS</title> The XGBoost model enables timely severity assessment, reduces risks of delayed diagnoses, and supports data-driven individualized interventions, demonstrating significant translational potential of this AI-driven approach for SLD management. </sec>

  PublisherDOI

• Research on Adaptive Control Systems for Building Shading Based on Decision Tree and Random Forest Classification Algorithms

  International Journal of Computer Science and Information Technology · 2024-09-13

  articleOpen accessSenior author

  Dynamic shading systems for buildings, due to their variable mechanisms and high adaptability, have the potential to reduce energy consumption in buildings significantly. However, the performance of dynamic shading is largely influenced by the control system employed. This study aims to explore an adaptive control method for building shading based on decision tree and random forest classification algorithms (machine learning), with the objective of minimizing energy demands related to lighting and cooling as much as possible. The adaptive control system model may independently modify the location of building shading in response to input environmental conditions, thereby achieving energy savings for both lighting and cooling. According to the findings, the automated control system model may reduce building energy consumption by 38% overall, which is close to the ideal level of energy conservation.

  PublisherOA PDFDOI

• Protective effects and mechanism of chemical- and plant-based selenocystine against cadmium-induced liver damage

  Journal of Hazardous Materials · 2024-02-16 · 20 citations

  article
  PublisherDOI

• A Study on the Diagnosis of the Working Conditions of a Traveling Beam Pumping Unit Based on Artificial Intelligence

  Frontiers in artificial intelligence and applications · 2024-11-21

  book-chapterOpen accessSenior author

  In order to solve the problem of low accuracy of traditional indicator diagram recognition methods, the research on condition diagnosis of beam pumping units based on artificial intelligence is proposed. Through the application of deep learning convolutional neural network in the field of image recognition, this paper proposes a convolutional neural network model based on LeNet, and realizes the automatic recognition of indicator diagrams. The model built by the research institute takes 15 common downhole working conditions of the pumping unit into consideration while simplifying the model structure, and introduces the Dropout layer and local response normalization layer to prevent the model from over fitting and improve the generalization ability of the model. The experimental results show that the model not only has a fast convergence rate, but also has an average diagnostic accuracy of 94.68% It meets the diagnostic accuracy requirements of pumping unit condition detection. Conclusion: This study provides a basis for the construction of pumping unit well condition intelligent monitoring and early warning system, which is of great significance to the construction of intelligent oilfield and the efficient production of oilfield.

  PublisherOA PDFDOI

Frequent coauthors

• Shaowen Wang

  University of Illinois Urbana-Champaign

  75 shared

• Anand Padmanabhan

  28 shared

• Wendy K. Tam Cho

  University of California, San Francisco

  22 shared

• Ian Gorton

  Northeastern University

  21 shared

• Hao Hu

  15 shared

• Jeff Terstriep

  National Center for Supercomputing Applications

  13 shared

• Hans H Stein

  University of Illinois Urbana-Champaign

  11 shared

• Liming Zhu

  Commonwealth Scientific and Industrial Research Organisation

  10 shared

Education

• ME, Computer Science

  Wuhan University

• PhD, Informatics

  University of Illinois at Urbana-Champaign

• BS, Computer Science

  Wuhan University

• MCS, Computer Science

  University of Iowa

Awards & honors

• Celebration of Excellence 2026
• Celebration of Excellence 2025
• Celebration of Excellence 2024
• Celebration of Excellence 2023
• Celebration of Excellence 2022