A5103226976- Bayesian Modeling and Causal Inference
- Robot Manipulation and Learning
- Data Mining Algorithms and Applications
- Machine Learning and Data Classification
- Text and Document Classification Technologies
- Rough Sets and Fuzzy Logic
- Anomaly Detection Techniques and Applications
- Human Pose and Action Recognition
- Inertial Sensor and Navigation
- Human-Automation Interaction and Safety
- Imbalanced Data Classification Techniques
- Robotics and Sensor-Based Localization
- Space Satellite Systems and Control
- Soft Robotics and Applications
- Speech Recognition and Synthesis
- Hand Gesture Recognition Systems
- Autonomous Vehicle Technology and Safety
- Astronomical Observations and Instrumentation
- Advanced Text Analysis Techniques
- Topic Modeling
- Hydrology and Watershed Management Studies
- Real-time simulation and control systems
- Vehicle Dynamics and Control Systems
- Spacecraft Design and Technology
- Lymphoma Diagnosis and Treatment
Beaumont Hospital, Royal Oak
2025
University of Wisconsin–Madison
2024
Toronto Metropolitan University
2017-2023
Google (United States)
2019-2023
Water Research Foundation
2018-2023
Florida State University
2023
Northeastern University
2023
Oregon State University
2023
Chesapeake Public Schools
2023
University of California, Berkeley
2020-2021
Naïve Bayes is one of the most efficient and effective inductive learning algorithms for machine data mining. Its competitive performance in classification surprising, because conditional independence assumption on which it based rarely true real-world applications. An open question is: what reason surprisingly good classification? In this paper, we propose a novel explanation Bayes. We show that, essentially, dependence distribution plays crucial role. Here means how local an attribute...
Article AbstractBackground: This study was designed to address the efficacy and tolerability of fluoxetine in patients with posttraumatic stress disorder (PTSD) as diagnosed using Structured Clinical Interview for DSM-IV Axis I Disorders Clinician-Administered PTSD Scale (CAPS). The patient population included both civilians combat veterans. Method: a double-blind, randomized, placebo-controlled conducted Europe, Israel, South Africa, primarily war-torn countries. Patients were predominantly...
Transfer learning and ensemble are the new trends for solving problem that training data test have different distributions. In this paper, we design an transfer framework to improve classification accuracy when insufficient. First, a weightedresampling method is proposed, which named TrResampling. each iteration, with heavy weights in source domain resampled, TrAdaBoost algorithm used adjust of target data. Second, three classic machine algorithms, namely, naive Bayes, decision tree, SVM, as...
Background Little is known about the effect of pharmacotherapy in prevention post-traumatic stress disorder (PTSD) relapse. Aims To assess efficacy and tolerability fluoxetine preventing PTSD Method This was a double-blind, randomised, placebo-controlled study. Following 12 weeks acute treatment, patients who responded were re-randomised continued 24-week relapse phase with ( n =69) or placebo =62). The primary assessment relapse, based on time to Results Patients fluoxetine/fluoxetine group...
Bayesian network classifiers have been widely used for classification problems. Given a fixed structure, parameters learning can take two different approaches: generative and discriminative learning. While parameter is more efficient, effective. In this paper, we propose simple, effective method, called Discriminative Frequency Estimate (DFE), which learns by discriminatively computing frequencies from data. Empirical studies show that the DFE algorithm integrates advantages of both...
Due to being fast, easy implement and relatively effective, some state-of-the-art naive Bayes text classifiers with the strong assumption of conditional independence among attributes, such as multinomial Bayes, complement one-versus-all-but-one model, have received a great deal attention from researchers in domain classification. In this article, we revisit these empirically compare their classification performance on large number widely used benchmark datasets. Then, propose locally...
We study the effectiveness of several techniques to personalize end-to-end speech models and improve recognition proper names relevant user. These differ in amounts user effort required provide supervision, are evaluated on how they impact performance. propose using keyword-dependent precision recall metrics measure vocabulary acquisition evaluate algorithms a dataset that we designed contain persons difficult recognize. Therefore, baseline rate for this is very low: 2.4%. A data synthesis...
Objective: This study was conducted to identify eye glance measures that are diagnostic of visual distraction. Background: Visual distraction degrades performance, but real-time have not been identified. Method: In a driving simulator, 14 participants responded lead vehicle braking at -2 or -2.7 m/s 2 periodically while reading varying number words (6-15 every 13 s) on peripheral displays (with diagonal eccentricities 24°, 43°, and 75°). Results: As the display eccentricity increased, total...
ABSTRACTObjectives: To build the structural model of pharmacokinetics for rosuvastatin and evaluate impact demographic characteristics including renal function on its pharmacokinetic parameters.Methods: A population analysis in healthy volunteers, subjects with dyslipidaemia, failure patients was performed using non-linear mixed-effects modelling a two-compartment simultaneous first- zero-order absorption. Demographic covariates, dyslipidaemic state were evaluated their parameters by...
We explore how high-speed robot arm motions can dynamically manipulate ropes and cables to vault over obstacles, knock objects from pedestals, weave between obstacles. In this paper, we propose a self-supervised learning framework that enables UR5 perform these three tasks. The finds 3D apex point for the arm, which, together with task-specific trajectory function, defines an arcing motion manipulates cable task varying obstacle target locations. function computes minimum-jerk are...
This paper examines the effectiveness of small star trackers for orbital estimation. Autonomous optical navigation has been used some time to provide local estimates parameters during close approach celestial bodies. These techniques have extensively on spacecraft dating back Voyager missions, but often rely long exposures and large instrument apertures. Using a hyperbolic Mars as reference mission, we present an EKF-based filter suitable nanosatellite missions. Observations its moons allow...
Accurate probability estimation generated by learning models is desirable in some practical applications, such as medical diagnosis. In this paper, we empirically study traditional decision-tree and their variants terms of estimation, measured conditional log likelihood (CLL). Furthermore, also compare decision tree with other kinds representative learning: Naive Bayes, Bayes tree, Bayesian network, K-nearest neighbors support vector machine respect to estimation. From our experiments, have...