A5076331761- Machine Learning in Materials Science
- Topic Modeling
- Advanced Text Analysis Techniques
- Computational Drug Discovery Methods
- Fault Detection and Control Systems
- Surface Modification and Superhydrophobicity
- Respiratory Support and Mechanisms
- Machine Learning and Data Classification
- Medical Imaging and Pathology Studies
- Chemistry and Chemical Engineering
- Sentiment Analysis and Opinion Mining
- Thermal Regulation in Medicine
- Advanced Sensor and Energy Harvesting Materials
- Aerogels and thermal insulation
- Sepsis Diagnosis and Treatment
- Nuclear Materials and Properties
- Hydrogen embrittlement and corrosion behaviors in metals
- Neural dynamics and brain function
- Adversarial Robustness in Machine Learning
- Risk and Safety Analysis
- Occupational Health and Safety Research
- Reinforcement Learning in Robotics
- Anomaly Detection Techniques and Applications
- Software Engineering Research
- Fusion materials and technologies
University of Wisconsin–Madison
2018-2022
University of Illinois Chicago
2021-2022
Wisconsin Institutes for Discovery
2018
Abstract Obtaining accurate estimates of machine learning model uncertainties on newly predicted data is essential for understanding the accuracy and whether its predictions can be trusted. A common approach to such uncertainty quantification estimate variance from an ensemble models, which are often generated by generally applicable bootstrap method. In this work, we demonstrate that direct standard deviation not but it simply calibrated dramatically improve accuracy. We effectiveness...
Abstract Irradiation increases the yield stress and embrittles light water reactor (LWR) pressure vessel steels. In this study, we demonstrate some of potential benefits risks using machine learning models to predict irradiation hardening extrapolated low flux, high fluence, extended life conditions. The training data included Variable for lower flux irradiations up an intermediate plus Belgian Reactor 2 Advanced Test 1 very irradiations, fluence. Notably, model predictions are superior...
Flash points of organic molecules play an important role in preventing flammability hazards and large databases measured values exist, although millions compounds remain unmeasured. To rapidly extend existing data to new many researchers have used quantitative structure-property relationship (QSPR) analysis effectively predict flash points. In recent years graph-based deep learning (GBDL) has emerged as a powerful alternative method traditional QSPR. this paper, GBDL models were implemented...
This paper proposes to study a fine-grained semantic novelty detection task, which can be illustrated with the following example. It is normal that person walks dog in park, but if someone says "A man walking chicken park", it novel. Given set of natural language descriptions scenes, we want identify novel scenes. We are not aware any existing work solves problem. Although or anomaly algorithms applicable, since they usually topic-based, perform poorly on our task. an effective model (called...
Much of the existing work on text novelty detection has been studied at topic level, i.e., identifying whether a document or sentence is novel not. Little done fine-grained semantic level (or contextual level). For example, given that we know Elon Musk CEO technology company, “Elon acted in sitcom The Big Bang Theory” and surprising because normally would not be an actor. Existing topic-based methods poorly this problem they do perform reasoning involving relations between named entities...
Designing Reinforcement Learning (RL) solutions for real-life problems remains a significant challenge. A major area of concern is safety. "Shielding" popular technique to enforce safety in RL by turning user-defined specifications into safe agent behavior. However, these methods either suffer from extreme learning delays, demand extensive human effort designing models and domains the problem, or require pre-computation. In this paper, we propose new permissibility-based framework deal with...
Abstract Background Acute lung injury and the acute respiratory distress syndrome are characterized by pulmonary inflammation, reduced endothelial barrier integrity filling of alveolar space with protein rich edema fluid infiltrating leukocytes. Animal models critical to uncovering pathologic mechanisms this devastating syndrome. Intravital imaging intact via two-photon intravital microscopy has proven a valuable method investigate in small rodent through characterization inflammatory cells...
Much of the existing work on text novelty detection has been studied at topic level, i.e., identifying whether a document or sentence is novel not. Little done fine-grained semantic level (or contextual level). For example, given that we know Elon Musk CEO technology company, "Elon acted in sitcom The Big Bang Theory" and surprising because normally would not be an actor. Existing topic-based methods poorly this problem they do perform reasoning involving relations between named entities...
Obtaining accurate estimates of machine learning model uncertainties on newly predicted data is essential for understanding the accuracy and whether its predictions can be trusted. A common approach to such uncertainty quantification estimate variance from an ensemble models, which are often generated by generally applicable bootstrap method. In this work, we demonstrate that direct standard deviation not propose a calibration method dramatically improve accuracy. We effectiveness both...