Scientific machine learning benchmarks

Deep learning has transformed the use of machine technologies for analysis large experimental datasets. In science, such datasets are typically generated by large-scale facilities, and focuses on identification patterns, trends anomalies to extract meaningful scientific insights from data. upcoming as Extreme Photonics Application Centre (EPAC) in UK or international Square Kilometre Array (SKA), rate data generation scale volumes will increasingly require more automated analysis. However, at present, identifying most appropriate algorithm any given dataset is a challenge due potential applicability many different frameworks, computer architectures models. Historically, modelling simulation high-performance computing systems, these issues have been addressed through benchmarking applications, algorithms architectures. Extending approach metrics application methods open, curated new both scientists scientists. Here, we introduce concept benchmarks science review existing approaches. As an example, describe SciMLBench suite benchmarks. Finding currently challenging, but being developed help.