A5064488664- Reinforcement Learning in Robotics
- Particle Detector Development and Performance
- Particle physics theoretical and experimental studies
- Radiation Detection and Scintillator Technologies
- Adversarial Robustness in Machine Learning
- Advanced Vision and Imaging
- Advanced Graph Neural Networks
- Distributed and Parallel Computing Systems
- Evolutionary Algorithms and Applications
- Software Engineering Research
- Fault Detection and Control Systems
- Data Stream Mining Techniques
- CCD and CMOS Imaging Sensors
- Medical Image Segmentation Techniques
- Advanced Data Compression Techniques
- Autonomous Vehicle Technology and Safety
- Advancements in PLL and VCO Technologies
- Adaptive Dynamic Programming Control
- Parallel Computing and Optimization Techniques
- Technology and Security Systems
- Traffic Prediction and Management Techniques
- Neural dynamics and brain function
- Human Mobility and Location-Based Analysis
- Bacillus and Francisella bacterial research
- 3D Shape Modeling and Analysis
The University of Texas at Austin
2024
Shenzhen University
2024
Stomatology Hospital
2024
University of Chinese Academy of Sciences
2021-2023
Institute of Automation
2021-2023
Chinese Academy of Sciences
2020-2023
Aerospace Information Research Institute
2023
Shandong Institute of Automation
2020-2023
Beijing Academy of Artificial Intelligence
2023
Fusion Academy
2022
Abstract Genetic engineering faces persistent challenges in achieving precise Deoxyribonucleic acid (DNA) cleavage, especially with the limitations associated current enzyme‐based methods, exemplified by issues CRISPR technologies. This study introduces a groundbreaking approach: utilizing reactive oxygen species (ROS) generated Multiphoton Absorption (MPA)‐excited Black Phosphorus Quantum Dots (BPQDs) under femto‐second laser irradiation. innovative method not only allows excitation lower...
Point cloud processing is rapidly expanding the applicable scenarios in industry. The surface normal a fundamental feature for various point tasks. Recently, deep supervised estimators outperform traditional estimation methods by adapting to dataset statistics. However, existing mainly adopt hand-crafted features or complicated networks borrowed from other tasks and make less effort design network models specifically problem. Instead of regressing vector directly, we propose simple estimate...
To design data visualizations that are easy to comprehend, we need understand how people with different interests read them. Computational models of predicting scanpaths on charts could complement empirical studies by offering estimates user performance inexpensively; however, previous have been limited gaze patterns and overlooked the effects tasks. Here, contribute Chartist, a computational model simulates users move their eyes extract information from chart in order perform analysis...
Recently, some challenging tasks in multi-agent systems have been solved by hierarchical reinforcement learning methods. Inspired the intra-level and inter-level coordination human nervous system, we propose a novel value decomposition framework HAVEN based on for fully cooperative problems. To address instability arising from concurrent optimization of policies between various levels agents, introduce dual mechanism inter-agent strategies designing reward functions two-level hierarchy. does...
Recent years have witnessed great potential in applying Deep Reinforcement Learning (DRL) various challenging applications, such as autonomous driving, nuclear fusion control, complex game playing, etc. However, recently researchers revealed that deep reinforcement learning models are vulnerable to adversarial attacks: malicious attackers can train policies tamper with the observations of a well-trained victim agent, latter which fails dramatically when faced an attack. Understanding and...
In this paper, we present the field programmable gate arrays (FPGA)-based framework intelligent FPGA data acquisition (IFDAQ), which is used for development of DAQ systems detectors in high-energy physics. The supports Xilinx and provides a collection IP cores written very high speed integrated circuit hardware description language, use common interconnect interface. core library offers functionality required full chain. consists Serializer/Deserializer (SERDES)-based time-to-digital...
COMPASS is a fixed-target experiment at the SPS CERN dedicated to study of hadron structure and spectroscopy. Since 2014, hardware event builder consisting nine custom designed FPGA-cards replaced previous online computers increasing compactness scalability DAQ. By buffering data, system exploits spill averages maximum on-spill data rate over whole cycle. From 2016, crosspoint switch connecting all involved high-speed links shall provide fully programmable topology thus simplifies...
In this paper we present the FPGA-based framework IFDAQ which is used for development of data acquisition systems detectors in high energy physics. The supports Xilinx FPGA and provides a collection IP cores written VHDL use common interconnect interface. core library offers functionality required full DAQ chain. consists SERDES-based TDC channels, an interface to multi-channel 80MS/s 10-bit ADC, transmission synchronization protocol between FPGA, event builder slow control. distributed...
Recent years have witnessed the great success of multi-agent systems (MAS). Value decomposition, which decom-poses joint action values into individual values, has been an important work in MAS. However, many value decomposition methods ignore coordination among different agents, leading to notorious "lazy agents" problem. To enhance MAS, this paper proposes HyperGraph CoNvo-lution MIX (HGCN-MIX), a method that incorporates hyper-graph convolution with decomposition. HGCN-MIX models agents as...
In the real world, many tasks require multiple agents to cooperate with each other under condition of local observations. To solve such problems, multi-agent reinforcement learning methods based on Centralized Training Decentralized Execution have been proposed. One representative class work is value decomposition, which decomposes global joint Q-value Q <inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">jt</inf> into individual Q-values...
The three-level air tank system is widely used in the industrial manufacture as gas storage device. Because of its features like time-delay and great inertia, usually it very difficult for traditional PID controller to achieve ideal control performance. Aiming at solving practical problems, this paper, we have proposed a constrained generalized predictive (GPC) algorithm that input amplitude, input-change-rate output are constrained. In algorithm, parameters identified by recursive least...
The unified communication framework (UCF) is a network protocol and field programmable gate array core for high-speed serial interfaces employed in data acquisition systems. It provides up to 64 different channels via single link. One channel reserved timing trigger information, whereas the other can be used slow control transmission. All except are bidirectional share bandwidth according assigned priority. distributes messages with fixed deterministic latency one direction. In this regard,...
PENeLOPE is a neutron lifetime measurement developed at the Technische Universität München and located Forschungs-Neutronenquelle Heinz Maier-Leibnitz (FRM II) aiming to achieve precision of 0.1 seconds. The detector for consists about 1250 Avalanche Photodiodes (APDs) with total active area 1225 cm2. decay proton electronics will be operated high electrostatic potential −30 kV magnetic field 0.6 T. This includes shaper, preamplifier, ADC FPGA cards. In addition, APDs cooled 77 K. are...
The Belle 2 experiment is undergoing an upgrade with the aim to perform most precise measurement of CP-violation using state-of-the-art detectors. silicon pixel detector innermost layer experiment. Because increased beam background and high detection efficiency detector, data rate estimated at 22 GB/s which 10 times higher than remaining detectors in We build a two stages FPGA-based read-out system controls prepares for online reduction. consists 48 modules AMC form factor use Xilinx...
Spatiotemporal predictive learning (ST-PL) is a hotspot with numerous applications, such as object movement and meteorological prediction. It aims at predicting the subsequent frames via observed sequences. However, inherent uncertainty among consecutive exacerbates difficulty in long-term To tackle increasing ambiguity during forecasting, we design CMS-LSTM to focus on context correlations multi-scale spatiotemporal flow details fine-grained locals, containing two elaborate designed blocks:...
The Unified Communication Framework is a unified network protocol and FPGA firmware for high speed serial interfaces employed in Data Acquisition systems. It provides up to 64 different communication channels via single link. One channel reserved timing trigger information whereas the other can be used slow control data transmission. All except are bidirectional share bandwidth according assigned priority. distributes messages with fixed deterministic latency one direction. In this regard...
The purely software based old data acquisition system of the COMPASS experiment at CERN is replaced by new hybrid FPGA-software system. two level FPGA subsystem takes over handling role. hardware built in a compact AMC form factor using Xilinx Virtex-6 VLX130T as processor. includes 15:1 link multiplexing and complete event building firmware. All high speed links are commutated 144×144 channels cross point switch which used to provide dynamic load balancing. designed throughput 1.6 GB/s...
Fine-Tuning Diffusion Models enable a wide range of personalized generation and editing applications on diverse visual modalities. While Low-Rank Adaptation (LoRA) accelerates the fine-tuning process, it still requires multiple reference images time-consuming training, which constrains its scalability for large-scale real-time applications. In this paper, we propose \textit{View Iterative Self-Attention Control (VisCtrl)} to tackle challenge. Specifically, VisCtrl is training-free method...
3D Gaussian Splatting demonstrates excellent quality and speed in novel view synthesis. Nevertheless, the huge file size of Gaussians presents challenges for transmission storage. Current works design compact models to replace substantial volume attributes Gaussians, along with intensive training distill information. These endeavors demand considerable time, presenting formidable hurdles practical deployment. To this end, we propose MesonGS, a codec post-training compression Gaussians....