A5114526252- Advanced Optimization Algorithms Research
- Machine Learning in Healthcare
- Generative Adversarial Networks and Image Synthesis
- Complexity and Algorithms in Graphs
- Computer Graphics and Visualization Techniques
- Numerical Methods and Algorithms
- Image and Signal Denoising Methods
- Polynomial and algebraic computation
- Topic Modeling
- Iterative Methods for Nonlinear Equations
- Cardiovascular Function and Risk Factors
- Domain Adaptation and Few-Shot Learning
- Cardiac Imaging and Diagnostics
- Reinforcement Learning in Robotics
- 3D Shape Modeling and Analysis
- Biomedical Text Mining and Ontologies
- Multimodal Machine Learning Applications
- Advanced Vision and Imaging
- Advanced Image and Video Retrieval Techniques
- Radar Systems and Signal Processing
- Augmented Reality Applications
- Artificial Intelligence in Healthcare and Education
- Optimization and Packing Problems
- Bayesian Modeling and Causal Inference
- Distributed Sensor Networks and Detection Algorithms
Yale University
2024-2025
University of Pennsylvania
2024-2025
University of Illinois Urbana-Champaign
2019-2023
Carnegie Mellon University
2021-2022
Xilinx (United States)
2020-2022
Princeton University
2017-2020
University of California, Berkeley
2017-2019
University of California, San Francisco
2017-2019
Purdue University West Lafayette
2006
Background: The ECG remains the most widely used diagnostic test for characterization of cardiac structure and electrical activity. We hypothesized that parallel advances in computing power, machine learning algorithms, availability large-scale data could substantially expand clinical inferences derived from while at same time preserving interpretability medical decision-making. Methods Results: identified 36 186 ECGs University California, San Francisco database would enable training models...
<title>Abstract</title> Recent advancements in large language models (LLMs) such as ChatGPT and LLaMA have hinted at their potential to revolutionize medical applications, yet application clinical settings often reveals limitations due a lack of specialized training on medical-specific data. In response this challenge, study introduces Me-LLaMA, novel LLM family that includes foundation – Me-LLaMA 13/70B, along with chat-enhanced versions 13/70B-chat, developed through continual pre-training...
Recent advancements in large language models (LLMs) show significant potential medical applications but are hindered by limited specialized knowledge. We present Me-LLaMA, a family of open-source LLMs integrating extensive domain-specific knowledge with robust instruction-following capabilities. Me-LLaMA is developed through continual pretraining and instruction tuning LLaMA2 using diverse biomedical clinical data sources (e.g., literature notes). evaluated on six text analysis tasks 12...
Abstract In real-world studies, the collected confounders may suffer from measurement error. Although mismeasurement of is typically unintentional (originating sources such as human oversight or imprecise machinery) deliberate also occurs and becoming increasingly more common. For example, in 2020 U.S. Census, noise was added to measurements assuage privacy concerns. Sensitive variables income age are often partially censored only known up a range values. settings, obtaining valid estimates...
We present a novel modular architecture for StarCraft II AI. The splits responsibilities between multiple modules that each control one aspect of the game, such as buildorder selection or tactics. A centralized scheduler reviews macros suggested by all and decides their order execution. An updater keeps track environment changes instantiates into series executable actions. Modules in this framework can be optimized independently jointly via human design, planning, reinforcement learning....
Virtual try-on methods aim to generate images of fashion models wearing arbitrary combinations garments. This is a challenging task because the generated image must appear realistic and accurately display interaction between Prior works produce that are filled with artifacts fail capture important visual details necessary for commercial applications. We propose Outfit Visualization Net (OVNet) these (e.g. buttons, shading, textures, hemlines, interactions garments) high quality...
Automated cardiac image interpretation has the potential to transform clinical practice in multiple ways including enabling low-cost serial assessment of function primary care and rural setting. We hypothesized that advances computer vision could enable building a fully automated, scalable analysis pipeline for echocardiogram (echo) interpretation. Our approach entailed: 1) preprocessing; 2) convolutional neural networks (CNN) view identification, segmentation, phasing cycle; 3)...
An increasing number of edge systems have large computational demands, stringent resource constraints, and end-to-end quality-driven goodness metrics. Architects embraced domain-specific accelerators to meet the demands such systems. We make case for research that shifts emphasis from systems, with a consequent shift evaluations using benchmarks are collections independent applications those testbeds full integrated describe extended reality (XR) as an exciting domain motivating research,...
We present Illinois Extended Reality testbed (ILLIXR), the first fully open-source XR system and research testbed. ILLIXR enables innovations with end-to-end co-designed hardware, compiler, OS, algorithms, driven by end-user perceived Quality-of-Experience (QoE) metrics. Using ILLIXR, we provide comprehensive quantitative analysis of performance, power, QoE for a complete its individual components. describe several implications our results that propel new directions in architecture, systems,...
Virtual dressing room applications help online shoppers visualize outfits. Such a system, to be commercially viable, must satisfy set of performance criteria. The system produce high quality images that faithfully preserve garment properties, allow users mix and match garments various types support human models varying in skin tone, hair color, body shape, so on. This paper describes POVNet, framework meets all these requirements (except shapes variations). Our uses warping methods together...
Recent large language models (LLMs) like ChatGPT and LLaMA have shown great promise in many AI applications. However, their performance on medical tasks is suboptimal can be further improved by training domain-specific datasets. This study introduces Me LLaMA, a LLM family including foundation - 13/70B chat-enhanced versions 13/70B-chat, developed through the continual pre-training instruction tuning of LLaMA2 using data. Our data suite for evaluation, includes large-scale dataset with 129B...
In this study, we analyze the risk of extreme value dependence in Chinese regional carbon emission markets. After filtering daily return data six markets China using a generalized autoregressive conditional heteroscedasticity (GARCH) model, obtain standardized residual series. Next, structures are captured by Copula function and Extreme Value theory (EVT). We report high peaks, heavy tails fluctuation aggregation logarithm series markets, as well significant dependent structures. There risks...
We consider the notions of (i) critical points, (ii) second-order (iii) local minima, and (iv) strict minima for multivariate polynomials. For each type point, as a function degree polynomial, we study complexity deciding (1) if given point is that type, (2) polynomial has type. Our results characterize these two questions all degrees left open by prior literature. main contributions reveal many turn out to be tractable cubic In particular, present an efficiently-checkable necessary...
We explore the power of semidefinite programming (SDP) for finding additive ɛ-approximate Nash equilibria in bimatrix games. introduce an SDP relaxation a quadratic formulation equilibrium problem and provide number valid inequalities to improve quality relaxation. If rank-1 solution this is found, then exact can be recovered. show that, strictly competitive game, our guaranteed return solution. propose two algorithms based on iterative linearization smooth nonconvex objective functions...
This report describes our approach for Phase 3 of the Real Robot Challenge. To solve cuboid manipulation tasks varying difficulty, we decompose each task into following primitives: moving fingers to grasp it, turning it on table minimize orientation error, and re-positioning goal position. We use model-based trajectory optimization control plan execute these primitives. These grasping, turning, primitives are sequenced with a state-machine that determines which primitive given current object...