A5107970374- Insect Pheromone Research and Control
- Advanced Chemical Sensor Technologies
- Advanced Topics in Algebra
- Plant Surface Properties and Treatments
- Theoretical and Computational Physics
- Advanced Data Compression Techniques
- Neural Networks and Applications
- Gaussian Processes and Bayesian Inference
- Stochastic processes and financial applications
- Fluid Dynamics and Heat Transfer
- Reinforcement Learning in Robotics
- Generative Adversarial Networks and Image Synthesis
- Neurobiology and Insect Physiology Research
- Distributed Control Multi-Agent Systems
- Stochastic processes and statistical mechanics
- Olfactory and Sensory Function Studies
- Robotic Path Planning Algorithms
- Species Distribution and Climate Change
- Model Reduction and Neural Networks
- Physics of Superconductivity and Magnetism
- Electrohydrodynamics and Fluid Dynamics
- Marine Toxins and Detection Methods
- Complex Network Analysis Techniques
- Insect Resistance and Genetics
- Superconducting Materials and Applications
RIKEN
2023-2025
Frankfurt Institute for Advanced Studies
2023-2025
Chinese University of Hong Kong, Shenzhen
2025
Swansea University
2025
China Agricultural University
2023-2024
National Natural Science Foundation of China
2023
Fudan University
2023
Embry–Riddle Aeronautical University
2018-2022
Abstract BACKGROUND The use of unmanned aerial vehicles (UAVs) for the application plant protection products (PPPs) in paddy fields is becoming increasingly prevalent worldwide. Despite its growing usage, UAV spraying rice pest control faces practical challenges, including limited canopy penetration, uneven deposition, and significant spray drift. This study investigated impact two tank‐mix adjuvants, Wonderful Rosin (Adjuvant‐1) Tiandun (Adjuvant‐2), at six volume concentrations, on...
The probability distribution effectively sampled by a complex Langevin process for theories with sign problem is not known priori and notoriously hard to understand. Diffusion models, class of generative AI, can learn distributions from data. In this contribution, we explore the ability diffusion models created process.
We develop diffusion models for simulating lattice gauge theories, where stochastic quantization is explicitly incorporated as a physical condition sampling. demonstrate the applicability of this novel sampler to U(1) theory in two spacetime dimensions and find that model trained at small inverse coupling constant can be extrapolated larger regions without encountering topological freezing problem. Additionally, employed sample configurations on different sizes requiring further training....
This article presents an olfactory-based navigation algorithm for using a mobile robot to locate odor source in turbulent flow environment. We analogize the localization as reinforcement learning problem. During plume tracing process, belief state partially observable Markov decision process model is adapted generate probability map that estimates possible locations, and hidden employed produce distribution premises propagation areas. Both estimations are fed robot, decision-making approach...
This paper presents an engineering-based chemical plume tracing (CPT) method for using on autonomous under-water vehicle (AUV) to locate a source in underwater environment with obstacles. Fundamental steps of the proposed are twofold. Firstly, estimated location is obtained by likelihood map, which generated based partially observable Markov decision process (POMDP). Secondly, after determined, A-star path planning algorithm used generate shortest toward target while avoiding Simulation...
We develop diffusion models for lattice gauge theories which build on the concept of stochastic quantization. This framework is applied to $U(1)$ theory in $1+1$ dimensions. show that a model trained at one small inverse coupling can be effectively transferred larger without encountering issues related topological freezing, i.e., generate configurations corresponding different couplings by introducing Boltzmann factors as physics conditions, while maintaining correct physical distributions...
To analyse how diffusion models learn correlations beyond Gaussian ones, we study the behaviour of higher-order cumulants, or connected n-point functions, under both forward and backward process. We derive explicit expressions for moment- cumulant-generating functionals, in terms distribution initial data properties It is shown analytically that during process cumulants are conserved without a drift, such as variance-expanding scheme, therefore endpoint maintains nontrivial correlations....
This paper presents implementations based on an extended Kalman filter (EKF) and unscented (UKF) for navigation of autonomous underwater vehicle (AUV). Maintaining accurate localization AUV is difficult because radio frequency signals, such as the global position system (GPS) are highly attenuated by water. To address this problem, proposes a new method inertial (INS) aided Doppler velocity log (DVL) short baseline (SBL). The presented EKF UKF fuse information from sensors to produce...
In this work, we establish a direct connection between generative diffusion models (DMs) and stochastic quantization (SQ). The DM is realized by approximating the reversal of process dictated Langevin equation, generating samples from prior distribution to effectively mimic target distribution. Using numerical simulations, demonstrate that can serve as global sampler for quantum lattice field configurations in two-dimensional $\phi^4$ theory. We DMs notably reduce autocorrelation times...
This study delves into the connection between machine learning and lattice field theory by linking generative diffusion models (DMs) with stochastic quantization, from a differential equation perspective. We show that DMs can be conceptualized reversing process driven Langevin equation, which then produces samples an initial distribution to approximate target distribution. In toy model, we highlight capability of learn effective actions. Furthermore, demonstrate its feasibility act as global...
The probability distribution effectively sampled by a complex Langevin process for theories with sign problem is not known priori and notoriously hard to understand. Diffusion models, class of generative AI, can learn distributions from data. In this contribution, we explore the ability diffusion models created process.
Diffusion models are currently the leading generative AI approach used for image generation in e.g. DALL-E and Stable Diffusion. In this talk we relate diffusion to stochastic quantisation field theory employ it generate configurations scalar fields on a two-dimensional lattice. We end with some speculations possible applications.