A5068431797- Evolutionary Algorithms and Applications
- Scientific Computing and Data Management
- Metaheuristic Optimization Algorithms Research
- Distributed and Parallel Computing Systems
- Advanced Proteomics Techniques and Applications
- Neural Networks and Applications
- Genomics and Phylogenetic Studies
- Reinforcement Learning in Robotics
- Advanced Memory and Neural Computing
- Scientific Measurement and Uncertainty Evaluation
- Remote-Sensing Image Classification
- Ferroelectric and Negative Capacitance Devices
- Anomaly Detection Techniques and Applications
- Autonomous Vehicle Technology and Safety
- Advanced Neural Network Applications
- African history and culture analysis
- Parallel Computing and Optimization Techniques
- Geographic Information Systems Studies
- Rangeland Management and Livestock Ecology
- Machine Learning in Bioinformatics
- Advanced Data Storage Technologies
- Machine Learning and Data Classification
- Advanced Multi-Objective Optimization Algorithms
- Domain Adaptation and Few-Shot Learning
- Land Rights and Reforms
Oak Ridge National Laboratory
2016-2024
Pennsylvania State University
2017
Institute for Advanced Study
2010-2014
George Mason University
1999-2014
Abstract In this Data Descriptor, we present county-level electricity outage estimates at 15-minute intervals from 2014 to 2022. By 2022 92% of customers in the 50 US States, Washington DC, and Puerto Rico are represented. These data have been produced by Environment for Analysis Geo-Located Energy Information (EAGLE-I TM ) , a geographic information system visualization platform created Oak Ridge National Laboratory map population experiencing outages every 15 minutes county level. Although...
The cost and efficiency of direct air capture (DAC) carbon dioxide (CO2) will be decisive in determining whether this technology can play a large role decarbonization. To probe the meteorological conditions on DAC we examine, at 1 × 1° resolution for continental United States (U.S.), impacts temperature, humidity, atmospheric pressure, CO2 concentration representative amine-based adsorption process. Spatial temporal variations pressure lead to strong available ambient across U.S. specific...
There are generally three types of scientific software users: users that solve problems using existing science tools, researchers explore new approaches by extending code, and educators teach students concepts. Python is a general-purpose programming language accessible to beginners, such as students, but also has rich ecosystem facilitates writing research software. Additionally, high-performance computing (HPC) resources become more readily available, support for parallel processing...
Deep learning has contributed to major advances in the prediction of protein structure from sequence, a fundamental problem structural bioinformatics. With predictions now approaching accuracy crystallographic experiments, and with accelerators like GPUs TPUs making inference using large models rapid, genome-level becomes an obvious aim. Leadership-class computing resources can be used perform genome-scale state-of-the-art deep models, providing wealth new data for systems biology...
Abstract Motivation Sphagnum-dominated peatlands store a substantial amount of terrestrial carbon. The genus is undersampled and under-studied. No experimental crystal structure from any Sphagnum species exists in the Protein Data Bank fewer than 200 Sphagnum-related genes have structural models available AlphaFold Structure Database. Tools resources are needed to help bridge these gaps, enable analysis other proteomes now made possible by accurate prediction. Results We present predicted...
Summary This paper presents a scalable object detection workflow for detecting objects, such as settlements, from remotely sensed (RS) imagery. We have successfully deployed this on Titan supercomputer and utilized it the task of mapping human settlement at country scale. The performance various stages in was analyzed before making operational. implemented strategies to address issues suboptimal resource utilization long‐tail effects due unbalanced image workload, data loss runtime failures,...
Architectural and hyperparameter design choices can influence deep-learner (DL) model fidelity but also be affected by malformed training validation data. However, practitioners may spend significant time refining layers hyperparameters before discovering that distorted data were impeding the progress. We found an evolutionary algorithm (EA) used to troubleshoot this kind of DL problem. An EA evaluated thousands configurations on Summit yielded no overall improvement in performance, which...
Neuromorphic computing technology continues to make strides in the development of new algorithms, devices, and materials. In addition, applications have begun emerge where neuromorphic shows promising results. However, numerous barriers further application remain. this work, we identify several science areas can either an immediate impact (within 1 3 years) or societal would be extremely high if technological addressed. We both opportunities hurdles for these areas. Finally, discuss future...
In many neuromorphic workflows, simulators play a vital role for important tasks such as training spiking neural networks, running neuroscience simulations, and designing, implementing, testing algorithms. Currently available cater to either workflows (e.g., NEST Brian2) or deep learning BindsNET). Problematically, the neuroscience-based are slow not very scalable, learning-based do support certain functionalities that typical of workloads synaptic delay). this paper, we address gap in...
Deep-learner hyper-parameters, such as kernel sizes, batch and learning rates, can significantly influence the quality of trained models. The state art for finding optimal hyper-parameters generally uses a brute force, grid search approach, random search, or Bayesian-based optimization among other techniques. We applied an evolutionary algorithm to optimize sizes convolutional neural network used detect settlements in satellite imagery. Usually layer are small - typically one, three, five...
Bloat is a common problem with Evolutionary Algorithms (EAs) that use variable length representation. By creating unnecessarily large individuals it results in longer EA runtimes and solutions are difficult to interpret. The causes of bloat still uncertain, but one theory suggests occurs when the phenotype (e.g. behaviors) parents not successfully inherited by their offspring. Noting similarity evolvability theory, which measures heritability fitness, we hypothesize reproductive operators...
Deep-learners have many hyper-parameters including learning rate, batch size, kernel size - all playing a significant role toward estimating high quality models. Discovering useful hyper-parameter guidelines is an active area of research, though the state art generally uses brute force, uniform grid approach or random search for finding ideal settings. We share preliminary results using alternative to deep learner tuning that evolutionary algorithm improve accuracy deep-learner models used...
Understanding and exploiting topographical data via standard machine learning techniques is challenging, mainly due to the large dynamic range of values present in elevation lack direct relationships between anthropogenic phenomena topography, when considering topographic-geology couplings, for instance. Here we consider first hurdle, range, an effort apply Convolutional Neural Network (CNN) approaches prediction human activity. CNN 3-D relies on normalization approaches, which only locally...
Abstract Asynchronous evolutionary algorithms are becoming increasingly popular as a means of making full use many processors while solving computationally expensive search and optimization problems. These excel at keeping large clusters fully utilized, but may sometimes inefficiently sample an excess fast‐evaluating solutions the expense higher‐quality, slow‐evaluating ones. We have previously introduced steady‐state parent selection strategy, SWEET (“Selection whilE EvaluaTing”), that...