Search for: All records

Ultra-intense laser–matter interactions are often difficult to predict from first principles because of the complexity of plasma processes and the many degrees of freedom relating to the laser and target parameters. An important approach to controlling and optimizing ultra-intense laser interactions involves gathering large datasets and using these data to train statistical and machine learning models. In this paper, we describe experimental efforts to accelerate electrons and protons to ∼MeV energies with this goal in mind. These experiments involve a 1 kHz repetition rate ultra-intense laser system with ∼10 mJ per shot, a peak intensity near 5 × 1018 W/cm2, and a “liquid leaf” target. Improvements to the data acquisition capabilities of this laser system greatly aided this investigation. Generally, we find that the trained models were very effective in controlling the numbers of MeV electrons ejected. The models were less successful at shifting the energy range of ejected electrons. Simultaneous control of the numbers of ∼MeV electrons and the energy range will be the subject of future experimentation using this platform. 

In this study, we consider three different machine‐learning methods—a three‐hidden‐layer neural network, support vector regression, and Gaussian process regression—and compare how well they can learn from a synthetic data set for proton acceleration in the Target Normal Sheath Acceleration regime. The synthetic data set was generated from a previously published theoretical model by Fuchs et al. 2005 that we modified. Once trained, these machine‐learning methods can assist with efforts to maximize the peak proton energy, or with the more general problem of configuring the laser system to produce a proton energy spectrum with desired characteristics. In our study, we focus on both the accuracy of the machine‐learning methods and the performance on one GPU including memory consumption. Although it is arguably the least sophisticated machine‐learning model we considered, support vector regression performed very well in our tests. 

ABSTRACT MotivationHere, we make available a second version of the BioTIME database, which compiles records of abundance estimates for species in sample events of ecological assemblages through time. The updated version expands version 1.0 of the database by doubling the number of studies and includes substantial additional curation to the taxonomic accuracy of the records, as well as the metadata. Moreover, we now provide an R package (BioTIMEr) to facilitate use of the database. Main Types of Variables IncludedThe database is composed of one main data table containing the abundance records and 11 metadata tables. The data are organised in a hierarchy of scales where 11,989,233 records are nested in 1,603,067 sample events, from 553,253 sampling locations, which are nested in 708 studies. A study is defined as a sampling methodology applied to an assemblage for a minimum of 2 years. Spatial Location and GrainSampling locations in BioTIME are distributed across the planet, including marine, terrestrial and freshwater realms. Spatial grain size and extent vary across studies depending on sampling methodology. We recommend gridding of sampling locations into areas of consistent size. Time Period and GrainThe earliest time series in BioTIME start in 1874, and the most recent records are from 2023. Temporal grain and duration vary across studies. We recommend doing sample‐level rarefaction to ensure consistent sampling effort through time before calculating any diversity metric. Major Taxa and Level of MeasurementThe database includes any eukaryotic taxa, with a combined total of 56,400 taxa. Software Formatcsv and. SQL.