A5029708613- Autonomous Vehicle Technology and Safety
- EEG and Brain-Computer Interfaces
- Epilepsy research and treatment
- Functional Brain Connectivity Studies
- Human Pose and Action Recognition
- Software Reliability and Analysis Research
- Traffic and Road Safety
- Neural dynamics and brain function
- 3D Shape Modeling and Analysis
- Fault Detection and Control Systems
- Risk and Safety Analysis
- Memory and Neural Mechanisms
- Face Recognition and Perception
- Human Motion and Animation
- Advanced Algorithms and Applications
- Robot Manipulation and Learning
- Spam and Phishing Detection
- Sentiment Analysis and Opinion Mining
- Image Processing and 3D Reconstruction
- Text and Document Classification Technologies
- Advanced Sensor and Control Systems
- Reinforcement Learning in Robotics
- Music and Audio Processing
- Time Series Analysis and Forecasting
- Anomaly Detection Techniques and Applications
University of California, Los Angeles
2019-2024
Tsinghua University
2024
UCLA Health
2022
Central South University
2005
Changsha University of Science and Technology
2004
Intracranially recorded interictal high-frequency oscillations have been proposed as a promising spatial biomarker of the epileptogenic zone. However, its visual verification is time-consuming and exhibits poor inter-rater reliability. Furthermore, no method currently available to distinguish generated from zone (epileptogenic oscillations) those other areas (non-epileptogenic oscillations). To address these issues, we constructed deep learning-based algorithm using chronic intracranial EEG...
Extracting meaning from a dynamic and variable flow of incoming information is major goal both natural artificial intelligence. Computer vision (CV) guided by deep learning (DL) has made significant strides in recognizing specific identity despite highly attributes. This the same challenge faced nervous system partially addressed concept cells-neurons exhibiting selective firing response to persons/places, described human medial temporal lobe (MTL) . Yet, access neurons representing...
We propose a novel framework, On-Demand MOtion Generation (ODMO), for generating realistic and diverse long-term 3D human motion sequences conditioned only on action types with an additional capability of customization. ODMO shows improvements over SOTA approaches all traditional evaluation metrics when evaluated three public datasets (HumanAct12, UESTC, MoCap). Furthermore, we provide both qualitative evaluations quantitative demonstrating several first-known customization capabilities...
Predicting future trajectories of road agents is a critical task for autonomous driving. Recent goal-based trajectory prediction methods, such as DenseTNT and PECNet [1], [2], have shown good performance on tasks public datasets. However, they usually require complicated goal-selection algorithms optimization. In this work, we propose KEMP, hierarchical end-to-end deep learning framework prediction. At the core our keyframe-based prediction, where keyframes are representative states that...
This study aimed to explore sensitive detection methods and deep learning (DL)-based classification for pathological high-frequency oscillations (HFOs).We analyzed interictal HFOs (80-500 Hz) in 15 children with medication-resistant focal epilepsy who underwent resection after chronic intracranial electroencephalogram via subdural grids. The were assessed using the short-term energy (STE) Montreal Neurological Institute (MNI) detectors examined features based on spike association...
Intelligent systems are increasingly integral to our daily lives, yet rare safety-critical events present significant latent threats their practical deployment. Addressing this challenge hinges on accurately predicting the probability of occurring within a given time step from current state, metric we define as 'criticality'. The complexity criticality arises extreme data imbalance caused by in high dimensional variables associated with events, refer curse rarity. Existing methods tend be...
Imitation learning (IL) is a well-known problem in the field of Markov decision process (MDP), where one given multiple demonstration trajectories generated by expert(s), and goal to replicate hidden expert-policies so that when MDP run independently, it generates close demonstrated ones. IL most useful tools used building versatile robots can learn from examples. This task becomes particularly challenging expert exhibits mixture behavior modes. Prior work has introduced latent variables...
To guarantee the safety and reliability of autonomous vehicle (AV) systems, corner cases play a crucial role in exploring system's behavior under rare challenging conditions within simulation environments. However, current approaches often fall short meeting diverse testing needs struggle to generalize novel, high-risk scenarios that closely mirror real-world conditions. tackle this challenge, we present AutoScenario, multimodal Large Language Model (LLM)-based framework for realistic case...
In advertising, identifying the content safety of web pages is a significant concern since advertisers do not want brands to be associated with threatening content. At same time, publishers would like maximize number on which they can place ads. Thus, fine balance must achieved while classifying in order satisfy both and publishers. this paper, we propose multimodal machine learning framework that fuses visual textual information from improve current predictions safety. The primary focus...
Imitation learning is the task of replicating expert policy from demonstrations, without access to a reward function. This becomes particularly challenging when exhibits mixture behaviors. Prior work has introduced latent variables model variations policy. However, our experiments show that existing works do not exhibit appropriate imitation individual modes. To tackle this problem, we adopt an encoder-free generative for behavior cloning (BC) accurately distinguish and imitate different...
We propose a novel framework, On-Demand MOtion Generation (ODMO), for generating realistic and diverse long-term 3D human motion sequences conditioned only on action types with an additional capability of customization. ODMO shows improvements over SOTA approaches all traditional evaluation metrics when evaluated three public datasets (HumanAct12, UESTC, MoCap). Furthermore, we provide both qualitative evaluations quantitative demonstrating several first-known customization capabilities...
Predicting future trajectories of road agents is a critical task for autonomous driving. Recent goal-based trajectory prediction methods, such as DenseTNT and PECNet, have shown good performance on tasks public datasets. However, they usually require complicated goal-selection algorithms optimization. In this work, we propose KEMP, hierarchical end-to-end deep learning framework prediction. At the core our keyframe-based prediction, where keyframes are representative states that trace out...
ABSTRACT Intracranially-recorded interictal high-frequency oscillations (HFOs) have been proposed as a promising spatial biomarker of the epileptogenic zone. However, visual verification HFOs is time-consuming and exhibits poor inter-rater reliability. Furthermore, no method currently available to distinguish generated from zone (epileptogenic HFOs: eHFOs) those other areas (non-epileptogenic non-eHFOs). To address these issues, we constructed deep learning (DL)-based algorithm using HFO...
Summary Extracting meaning from a dynamic and variable flow of incoming information is major goal both natural artificial intelligence. Computer vision (CV) guided by deep learning (DL) has made significant strides in recognizing specific identity despite highly attributes 1,2 . This the same challenge faced nervous system partially addressed concept cells—neurons exhibiting selective firing response to persons/places, described human medial temporal lobe (MTL) 3–6 Yet, access neurons...