A5028849793- Autism Spectrum Disorder Research
- Radiomics and Machine Learning in Medical Imaging
- Child Development and Digital Technology
- AI in cancer detection
- Gaze Tracking and Assistive Technology
- Genetics and Neurodevelopmental Disorders
- Cloud Computing and Resource Management
- Glioma Diagnosis and Treatment
- Cutaneous Melanoma Detection and Management
- Digital Imaging for Blood Diseases
- Glaucoma and retinal disorders
- Cognitive Abilities and Testing
- Child and Animal Learning Development
- Neuroscience and Neuropharmacology Research
- Infrared Thermography in Medicine
- Natural Language Processing Techniques
- Machine Learning and Data Classification
- Visual Attention and Saliency Detection
- Virology and Viral Diseases
- Digital Accessibility for Disabilities
- Semantic Web and Ontologies
- Risk and Portfolio Optimization
- Face Recognition and Perception
- Epilepsy research and treatment
- Neural Networks and Applications
Microsoft (United States)
2022-2025
Microsoft Research (United Kingdom)
2023-2024
University of Washington
2018-2023
Yale University
2016-2023
Seattle Children's Hospital
2018-2023
University of Liverpool
2023
Queen's University
2023
Swedish Collegium for Advanced Study
2023
Karolinska Institutet
2023
Uppsala University
2023
AutoGen is an open-source framework that allows developers to build LLM applications via multiple agents can converse with each other accomplish tasks. are customizable, conversable, and operate in various modes employ combinations of LLMs, human inputs, tools. Using AutoGen, also flexibly define agent interaction behaviors. Both natural language computer code be used program flexible conversation patterns for different applications. serves as a generic infrastructure diverse complexities...
Eye tracking (ET) is a powerful methodology for studying attentional processes through quantification of eye movements. The precision, usability, and cost-effectiveness ET render it promising platform developing biomarkers use in clinical trials autism spectrum disorder (ASD).
In this paper, we introduce an end-to-end machine learning-based system for classifying autism spectrum disorder (ASD) using facial attributes such as expressions, action units, arousal, and valence. Our classifies ASD representations of different from convolutional neural networks, which are trained on images in the wild. experimental results show that used our statistically significant improve sensitivity, specificity, F1 score classification by a large margin. particular, addition...
Supply chain operations traditionally involve a variety of complex decision making problems. Over the last few decades, supply chains greatly benefited from advances in computation, which allowed transition manual processing to automation and cost-effective optimization. Nonetheless, business operators still need spend substantial efforts explaining interpreting optimization outcomes stakeholders. Motivated by recent Large Language Models (LLMs), we study how this disruptive technology can...
Problem definition: Cloud computing is a multibillion-dollar business that draws substantial capital investments from large companies such as Amazon, Microsoft, and Google. Large cloud providers need to accommodate the growing demand for resources while avoiding unnecessary overprovisioning of hardware operational costs. The underlying decision processes are challenging, they involve long-term infrastructure under future uncertainty. In this paper, we introduce server deployment problem. One...
Recent AI advancements, such as OpenAI's new models, are transforming LLMs into LRMs (Large Reasoning Models) that perform reasoning during inference, taking extra time and compute for higher-quality outputs. We aim to uncover the algorithmic framework training LRMs. Methods like self-consistency, PRM, AlphaZero suggest guided search. ask: what is simplest, most scalable way enable search in LLMs? propose a post-training called Reinforcement Learning via Self-Play (RLSP). RLSP involves three...
Diagnosing diseases through histopathology whole slide images (WSIs) is fundamental in modern pathology but challenged by the gigapixel scale and complexity of WSIs. Trained histopathologists overcome this challenge navigating WSI, looking for relevant patches, taking notes, compiling them to produce a final holistic diagnostic. Traditional AI approaches, such as multiple instance learning transformer-based models, fail short holistic, iterative, multi-scale diagnostic procedure, limiting...
Recent learned cardinality estimation (CE) models are vulnerable when query predicates or the underlying datasets drift from what were trained upon. We propose a system Warper that accelerates model adaptation to drifts; generates additional queries limited examples available new workload and carefully picks which use update CE model. show can be used adapt different including ones support over single tables join expressions. Experiments with drifts suggest has small computational cost...
Abstract Aims To compare different patterns of memory impairment in patients with two subtypes mesial temporal lobe epilepsy (MTLE) and healthy controls. Methods Thirty‐five controls 41 MTLE were recruited, which 25 diagnosed as hippocampal sclerosis (HS‐MTLE), the rest 16 lesion‐negative (MRI‐neg MTLE). Participants completed Wechsler assessment a short‐term game on an automated computer‐based platform eye tracker. Results Both MRI‐neg HS‐MTLE groups took longer time to complete than ( p...
In their expanding role as tutors, home and healthcare assistants, robots must effectively interact with individuals of varying ability temperament. Indeed, deploying in long-term social engagements will almost certainly require to reliably detect adapt changes the demeanor partners promote trust more productive collaboration. However, recognition emotional state typically relies on interpretation very subtle cues, often from one person next. addition, while facial expressions, body posture...
Most studies of executive function (EF) in Autism Spectrum Disorder (ASD) focus on cognitive information processing, emphasizing less the social interaction deficits core to ASD. We designed a mobile game that uses and nonsocial stimuli assess children's EF skills. The comprised three components involving different skills: flexibility (shifting/inference), inhibitory control, short-term memory. By recruiting 65 children with without ASD play game, we investigated potential such platforms for...
Abstract The Selective Social Attention (SSA) task is a brief eye‐tracking involving experimental conditions varying along socio‐communicative axes. Traditionally the SSA has been used to probe socially‐specific attentional patterns in infants and toddlers who develop autism spectrum disorder (ASD). This current work extends these findings preschool school‐age children. Children 4‐ 12‐years‐old with ASD ( N = 23) typically‐developing comparison group (TD; 25) completed as well standardized...
In this paper, we investigate the intersection of large generative AI models and cloud-native computing architectures. Recent such as ChatGPT, while revolutionary in their capabilities, face challenges like escalating costs demand for high-end GPUs. Drawing analogies between large-model-as-a-service (LMaaS) cloud database-as-a-service (DBaaS), describe an AI-native paradigm that harnesses power both technologies (e.g., multi-tenancy serverless computing) advanced machine learning runtime...
Background: Looking pattern differences are shown to separate individuals with Autism Spectrum Disorder (ASD) and Typically Developing (TD) controls. Recent studies have that, in children ASD, these patterns change intellectual social impairments, suggesting that of attention provide indices clinically meaningful variation ASD. Method: We conducted a naturalistic study ASD (n = 55) typical development (TD, n 32). A battery eye-tracking video stimuli was used the study, including Activity...
Detection of melanocytes serves as a critical prerequisite in assessing melanocytic growth patterns when diagnosing melanoma and its precursor lesions on skin biopsy specimens. However, this detection is challenging due to the visual similarity other cells routine Hematoxylin Eosin (H&E) stained images, leading failure current nuclei methods. Stains such Sox10 can mark melanocytes, but they require an additional step expense thus are not regularly used clinical practice. To address these...
This paper modifies the DBSCAN algorithm to identify fixations and saccades. method combines advantages from dispersion-based algorithms, such as resilience noise intuitive fixational structure, velocity-based ability deal appropriately with smooth pursuit (SP) movements.
Generalized linear models (GLMs) are well-established tools for regression and classification widely applied across the sciences, economics, business, finance.Owing to their convex loss, they easy efficient fit.Moreover, relatively interpret because of well-defined noise distributions point-wise nonlinearities.
Objective: To explore quantitative measurements of the visual attention and neuroelectrophysiological relevance memory deficits in temporal lobe epilepsy (TLE) by eye tracking electroencephalography (EEG). Methods: Thirty-four TLE patients twenty-eight healthy controls were invited to complete neurobehavioral assessments, cognitive oculomotor tasks, 24-h video EEG (VEEG) recordings using an automated computer-based assessment platform with tracker. Visit counts, visit time, time first...
Melanoma is the third most common type of skin cancer and responsible for deaths. A diagnosis melanoma made by visual interpretation tissue sections a pathologist, challenging task given complexity breadth melanocytic lesions subjective nature biopsy interpretation. We leverage advances in computer vision to aid segmenting potential regions on digital images whole slide biopsies. In this study, we demonstrate Mask-R-CNN-based segmentation framework such purpose. To alleviate cost data...
Invasive melanoma, a common type of skin cancer, is considered one the deadliest. Pathologists routinely evaluate melanocytic lesions to determine amount atypia, and if lesion represents an invasive its stage. However, due complicated nature these assessments, inter- intra-observer variability among pathologists in their interpretation are very common. Machine-learning techniques have shown impressive robust performance on various tasks including healthcare. In this work, we study potential...
Eye tracking has become a powerful tool in the study of autism spectrum disorder (ASD). Current, large-scale efforts aim to identify specific eye-tracking stimuli be used as biomarkers for ASD, with intention informing diagnostic process, monitoring therapeutic response, predicting outcomes, or identifying subgroups spectrum. However, there are hundreds candidate experimental paradigms, each which contains dozens even individual stimuli. Each is associated an array potential derived outcome...
Amorphous calcifications noted on mammograms (i.e., small and indistinct that are difficult to characterize) associated with high diagnostic uncertainty, often leading biopsies. Yet, only 20% of biopsied amorphous cancer. We present a quantitative approach for distinguishing between benign actionable (high-risk malignant) using combination local textures, global spatial relationships, interpretable handcrafted expert features.Our was trained validated set 168 2D full-field digital...
PURPOSE Machine Learning Package for Cancer Diagnosis (MLCD) is the result of a National Institutes Health/National Institute (NIH/NCI)-sponsored project developing unified software package from state-of-the-art breast cancer biopsy diagnosis and machine learning algorithms that can improve quality both clinical practice ongoing research. METHODS Whole-slide images 240 well-characterized cases, initially assembled under R01 CA140560, were used training models. This based on methodology...
Far infrared thermography, which can be used to detect thermal radiation emitted by humans, has been physical disease, physiological changes relating emotion, and polygraph testing, but not for eye tracking. However, because the surface temperature of cornea is colder than limbus, it theoretically possible track corneal movements through imaging. To explore feasibility tracking, we invited 10 adults tracked their with passive imaging at 60 Hz. We combined shape models eyes intensity...