A5062396178- Ferroelectric and Negative Capacitance Devices
- Fire effects on ecosystems
- Speech Recognition and Synthesis
- Landslides and related hazards
- Advanced Memory and Neural Computing
- Topic Modeling
- Machine Learning and Data Classification
- Natural Language Processing Techniques
- demographic modeling and climate adaptation
- Soil Geostatistics and Mapping
- Forest ecology and management
- Data Stream Mining Techniques
- High-Velocity Impact and Material Behavior
- Neural Networks and Applications
- Energy Load and Power Forecasting
- Advanced Bandit Algorithms Research
- Engineering Applied Research
- Neural Networks and Reservoir Computing
- Gaussian Processes and Bayesian Inference
- Hydrology and Watershed Management Studies
- Climate variability and models
- Information and Cyber Security
- Magnetic properties of thin films
- Atmospheric and Environmental Gas Dynamics
- GNSS positioning and interference
Los Alamos National Laboratory
2018-2024
Virginia Commonwealth University
2015-2018
The need for increasingly powerful computing hardware has spawned many ideas stipulating, primarily, the replacement of traditional transistors with alternate "switches" that dissipate miniscule amounts energy when they switch and provide additional functionality are beneficial information processing. An interesting idea emerged recently is notion using two-phase (piezoelectric/magnetostrictive) multiferroic nanomagnets bistable (or multi-stable) magnetization states to encode digital...
Wildfires present intricate challenges for prediction, necessitating the use of sophisticated machine learning techniques effective modeling\cite{jain2020review}. In our research, we conducted a thorough assessment various algorithms both classification and regression tasks relevant to predicting wildfires. We found that classifying different types or stages wildfires, XGBoost model outperformed others in terms accuracy robustness. Meanwhile, Random Forest showed superior results extent...
Almost all scientific data have uncertainties originating from different sources. Gaussian process regression (GPR) models are a natural way to model with Gaussian-distributed uncertainties. GPR also has the benefit of reducing I/O bandwidth and storage requirements for large simulations. However, reconstruction suffers high computation complexity. To make situation worse, classic approaches visualizing uncertainties, like probabilistic marching cubes, computationally very expensive,...
We propose a reconfigurable bit comparator implemented with nanowire spin valve whose two contacts are magnetostrictive bistable magnetization. Reference and input bits "written" into the magnetization states of electrically generated strain spin-valve's resistance is lowered if they match. Multiple comparators can be interfaced in parallel magneto-tunneling junction to determine an N-bit stream matches reference by bit. The system robust against thermal noise at room temperature 16-bit...
Straintronics is an extraordinarily energy-efficient novel hardware paradigm for digital computing and signal processing. The central idea to build the basic binary switch with a nanoscale multiferroic consisting of magnetostrictive layer elastically coupled piezoelectric layer. A tiny voltage few mV generates sufficient strain in magnetization nanomagnet memory or logic operation, while dissipating aJ energy. Low density processors employing straintronics operating at slow clock speeds ∼100...
The advent of large language models (LLMs) has significantly advanced the field code translation, enabling automated translation between programming languages. However, these often struggle with complex tasks due to inadequate contextual understanding. This paper introduces a novel approach that enhances through Few-Shot Learning, augmented retrieval-based techniques. By leveraging repository existing translations, we dynamically retrieve most relevant examples guide model in translating new...
Prescribed fires are an important part of forest stewardship in Western North America, understanding prescribed burn behavior is because if done incorrectly can result unintended burned land as well harm to humans and the environment. We looked at ensemble datasets from QUIC-Fire, a fire-atmospheric modeling tool, compared various machine learning models effectiveness predicting outcome variables, such area inside outside control boundary, fire was safe or unsafe. It found that out tested...