{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2025,9,23]],"date-time":"2025-09-23T13:51:38Z","timestamp":1758635498568,"version":"3.37.3"},"reference-count":45,"publisher":"Springer Science and Business Media LLC","issue":"1","license":[{"start":{"date-parts":[[2021,7,5]],"date-time":"2021-07-05T00:00:00Z","timestamp":1625443200000},"content-version":"tdm","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0"},{"start":{"date-parts":[[2021,7,5]],"date-time":"2021-07-05T00:00:00Z","timestamp":1625443200000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0"}],"funder":[{"name":"Alma Mater Studiorum - Universit\u00e0 di Bologna"}],"content-domain":{"domain":["link.springer.com"],"crossmark-restriction":false},"short-container-title":["Comput Softw Big Sci"],"published-print":{"date-parts":[[2021,12]]},"abstract":"<jats:title>Abstract<\/jats:title><jats:p>Machine Learning (ML) will play a significant role in the success of the upcoming High-Luminosity LHC (HL-LHC) program at CERN. An unprecedented amount of data at the exascale will be collected by LHC experiments in the next decade, and this effort will require novel approaches to train and use ML models. In this paper, we discuss a Machine Learning as a Service pipeline for HEP (MLaaS4HEP) which provides three independent layers: a data streaming layer to read High-Energy Physics (HEP) data in their native ROOT data format; a data training layer to train ML models using distributed ROOT files; a data inference layer to serve predictions using pre-trained ML models via HTTP protocol. Such modular design opens up the possibility to train data at large scale by reading ROOT files from remote storage facilities, e.g., World-Wide LHC Computing Grid (WLCG) infrastructure, and feed the data to the user\u2019s favorite ML framework. The inference layer implemented as TensorFlow as a Service (TFaaS) may provide an easy access to pre-trained ML models in existing infrastructure and applications inside or outside of the HEP domain. In particular, we demonstrate the usage of the MLaaS4HEP architecture for a physics use-case, namely, the <jats:inline-formula><jats:alternatives><jats:tex-math>$$t{\\bar{t}}$$<\/jats:tex-math><mml:math xmlns:mml=\"http:\/\/www.w3.org\/1998\/Math\/MathML\">\n                  <mml:mrow>\n                    <mml:mi>t<\/mml:mi>\n                    <mml:mover>\n                      <mml:mrow>\n                        <mml:mi>t<\/mml:mi>\n                      <\/mml:mrow>\n                      <mml:mrow>\n                        <mml:mo>\u00af<\/mml:mo>\n                      <\/mml:mrow>\n                    <\/mml:mover>\n                  <\/mml:mrow>\n                <\/mml:math><\/jats:alternatives><\/jats:inline-formula> Higgs analysis in CMS originally performed using custom made Ntuples. We provide details on the training of the ML model using distributed ROOT files, discuss the performance of the MLaaS and TFaaS approaches for the selected physics analysis, and compare the results with traditional methods.<\/jats:p>","DOI":"10.1007\/s41781-021-00061-3","type":"journal-article","created":{"date-parts":[[2021,7,5]],"date-time":"2021-07-05T19:02:55Z","timestamp":1625511775000},"update-policy":"https:\/\/doi.org\/10.1007\/springer_crossmark_policy","source":"Crossref","is-referenced-by-count":7,"title":["MLaaS4HEP: Machine Learning as a Service for HEP"],"prefix":"10.1007","volume":"5","author":[{"ORCID":"https:\/\/orcid.org\/0000-0003-0667-069X","authenticated-orcid":false,"given":"Valentin","family":"Kuznetsov","sequence":"first","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"ORCID":"https:\/\/orcid.org\/0000-0003-3539-4313","authenticated-orcid":false,"given":"Luca","family":"Giommi","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-0835-9574","authenticated-orcid":false,"given":"Daniele","family":"Bonacorsi","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]}],"member":"297","published-online":{"date-parts":[[2021,7,5]]},"reference":[{"key":"61_CR1","unstructured":"Albertsson K et al (2018) Machine Learning in High Energy Physics Community White Paper. arXiv:1807.02876 [physics.comp-ph]"},{"key":"61_CR2","unstructured":"A Living Review of Machine Learning for Particle Physics. https:\/\/iml-wg.github.io\/HEPML-LivingReview\/"},{"key":"61_CR3","unstructured":"A modular scientific toolkit used in HEP for analysis and as a data-storage format. https:\/\/root.cern.ch"},{"key":"61_CR4","unstructured":"Comma Separated Values (CSV) data-format. https:\/\/www.wikiwand.com\/en\/Comma-separated_values"},{"key":"61_CR5","unstructured":"Scientific package to represent data as multi-dimensional arrays. http:\/\/www.numpy.org"},{"key":"61_CR6","unstructured":"Hierarchical Data Format. https:\/\/www.wikiwand.com\/en\/Hierarchical_Data_Format"},{"key":"61_CR7","unstructured":"Higgs Boson Machine Learning Challenge used by ATLAS experiment to identify Higgs boson. https:\/\/www.kaggle.com\/c\/higgs-boson"},{"key":"61_CR8","unstructured":"High Energy Physics particle tracking in CERN detectors. https:\/\/www.kaggle.com\/c\/trackml-particle-identification"},{"key":"61_CR9","unstructured":"Flavours of Physics: Finding $$\\tau \\rightarrow \\mu \\mu \\mu$$. https:\/\/www.kaggle.com\/c\/flavours-of-physics\/data"},{"key":"61_CR10","doi-asserted-by":"publisher","unstructured":"Kuznetsov V, Giommi L (2018) MLaaS4HEP is set of MLaaS components for HEP. DOI: https:\/\/doi.org\/10.5281\/zenodo.1481785, https:\/\/github.com\/vkuznet\/MLaaS4HEP","DOI":"10.5281\/zenodo.1481785"},{"key":"61_CR11","doi-asserted-by":"publisher","unstructured":"Kuznetsov V (2018) TensorFlow as a Service. https:\/\/doi.org\/10.5281\/zenodo.1308049, http:\/\/github.com\/vkuznet\/TFaaS","DOI":"10.5281\/zenodo.1308049"},{"key":"61_CR12","doi-asserted-by":"publisher","unstructured":"Yao Y et al (2017) Complexity vs. performance: empirical analysis of machine learning as a service. In: Proceedings of the 2017 Internet Measurement Conference, pp 384-397. https:\/\/doi.org\/10.1145\/3131365.3131372","DOI":"10.1145\/3131365.3131372"},{"key":"61_CR13","unstructured":"CERN Openlab. https:\/\/home.cern\/science\/computing\/cern-openlab"},{"key":"61_CR14","unstructured":"A package for machine learning inference in FPGAs. https:\/\/hls-fpga-machine-learning.github.io\/hls4ml"},{"key":"61_CR15","unstructured":"Services for Optimal Network Inference on Coprocessors. https:\/\/github.com\/hls-fpga-machine-learning\/SonicCMS"},{"key":"61_CR16","unstructured":"Machine Learning Pipelines for High Energy Physics Using Apache Spark with BigDL and Analytics Zoo. https:\/\/db-blog.web.cern.ch\/blog\/luca-canali\/machine-learning-pipelines-high-energy-physics-using-apache-spark-bigdl"},{"key":"61_CR17","doi-asserted-by":"publisher","unstructured":"CMS Collaboration (2020) A deep neural network to search for new long-lived particles decaying to jets. Mach. Learn.: Sci. Technol. 1: 035012. DOI: https:\/\/doi.org\/10.1088\/2632-2153\/ab9023","DOI":"10.1088\/2632-2153\/ab9023"},{"key":"61_CR18","unstructured":"DIANA-HEP Scikit-hep uproot library. Minimalist ROOT I\/O in pure Python and NumPy. https:\/\/github.com\/scikit-hep\/uproot"},{"key":"61_CR19","unstructured":"A high performance, scalable fault tolerant access to data repositories of many kinds. http:\/\/xrootd.org"},{"key":"61_CR20","unstructured":"PyTorch AI library. https:\/\/www.pytorch.org"},{"key":"61_CR21","unstructured":"Keras AI library. https:\/\/keras.io"},{"key":"61_CR22","unstructured":"Tensor Flow AI library. http:\/\/www.tensorflow.org"},{"key":"61_CR23","unstructured":"Fast AI library. https:\/\/www.fast.ai"},{"key":"61_CR24","unstructured":"Kuznetsov V (2018) Machine Learning as a Service for HEP. arXiv:1811.04492 [hep-ex]"},{"key":"61_CR25","unstructured":"An umbrella organization for bringing state-of-the art for HEP experiments. http:\/\/diana-hep.org"},{"key":"61_CR26","unstructured":"Python Generator concept. https:\/\/wiki.python.org\/moin\/Generators"},{"key":"61_CR27","unstructured":"G\u00e9ron A (2019) Hands-On Machine Learning with Scikit-Learn, Keras, and TensorFlow. O\u2019Reilly, ISBN: 9781492032632"},{"key":"61_CR28","unstructured":"Stochastic Gradient Descent. https:\/\/scikit-learn.org\/stable\/modules\/sgd.html"},{"key":"61_CR29","unstructured":"DNN\/TensorFlow interface for CMSSW. https:\/\/github.com\/mharrend\/CMSSW-DNN"},{"key":"61_CR30","doi-asserted-by":"publisher","unstructured":"ATLAS Lightweight Trained Neural Network. https:\/\/doi.org\/10.5281\/zenodo.4299114, https:\/\/github.com\/lwtnn\/lwtnn","DOI":"10.5281\/zenodo.4299114"},{"key":"61_CR31","unstructured":"Go programming language. http:\/\/www.golang.org"},{"key":"61_CR32","unstructured":"CMS Collaboration (2016) Search for $${{t\\bar t}}$$H production in the H $$\\rightarrow$$$$b\\overline b$$ decay channel with $$\\sqrt{s} =$$ 13 TeV pp collisions at the CMS experiment. CMS PAS HIG-16-004"},{"issue":"2\u20133","key":"61_CR33","doi-asserted-by":"publisher","first-page":"77","DOI":"10.1393\/ncc\/i2020-20077-4","volume":"43","author":"F Iemmi","year":"2020","unstructured":"Iemmi F (2020) $${t\\bar{t}H}$$ associated production in the all-jets final state with the CMS experiment. Nuovo Cim. C 43(2\u20133):77. https:\/\/doi.org\/10.1393\/ncc\/i2020-20077-4","journal-title":"Nuovo Cim. C"},{"issue":"1","key":"61_CR34","doi-asserted-by":"publisher","first-page":"1","DOI":"10.1016\/j.physletb.2012.08.020","volume":"716","author":"ATLAS Collaboration","year":"2012","unstructured":"Collaboration ATLAS (2012) Observation of a New Particle in the Search for the Standard Model Higgs Boson with the ATLAS Detector at the LHC. Phys Lett B 716(1):1\u201329. https:\/\/doi.org\/10.1016\/j.physletb.2012.08.020","journal-title":"Phys Lett B"},{"issue":"1","key":"61_CR35","doi-asserted-by":"publisher","first-page":"30","DOI":"10.1016\/j.physletb.2012.08.021","volume":"716","author":"CMS Collaboration","year":"2012","unstructured":"Collaboration CMS (2012) Observation of a New Boson at a Mass of 125 GeV with the CMS Experiment at the LHC. Phys Lett B 716(1):30\u201361. https:\/\/doi.org\/10.1016\/j.physletb.2012.08.021","journal-title":"Phys Lett B"},{"key":"61_CR36","unstructured":"Hoecker A, Speckmayer P, Stelzer J, Therhaag J, von Toerne E, Voss H (2007) TMVA: Toolkit for Multivariate Data Analysis. arXiv:physics\/0703039 [physics.data-an]"},{"key":"61_CR37","unstructured":"ProtoBuffer library. https:\/\/github.com\/protocolbuffers\/protobuf"},{"key":"61_CR38","doi-asserted-by":"publisher","unstructured":"Tracolli M et al (2020) Using DODAS as deployment manager for smart caching of CMS data management system. J. Phys.: Conf. Ser. 1525: 012057. https:\/\/doi.org\/10.1088\/1742-6596\/1525\/1\/012057","DOI":"10.1088\/1742-6596\/1525\/1\/012057"},{"key":"61_CR39","doi-asserted-by":"crossref","unstructured":"Ben-Nun T, Hoefler T (2018) Demystifying Parallel and Distributed Deep Learning: An In-Depth Concurrency Analysis. arXiv:1802.09941 [cs.LG]","DOI":"10.1145\/3320060"},{"key":"61_CR40","unstructured":"Scalable Analytics framework in Python. https:\/\/dask.org"},{"key":"61_CR41","unstructured":"Distributed Keras framework. https:\/\/github.com\/cerndb\/dist-keras"},{"key":"61_CR42","unstructured":"Anderson D et al (2017) An MPI-Based Python Framework for Distributed Training with Keras. arXiv:1712.05878 [cs.DC]"},{"key":"61_CR43","unstructured":"An open source platform for the machine learning life cycle. https:\/\/www.mlflow.org"},{"key":"61_CR44","unstructured":"Open Neural Network Exchange format. http:\/\/www.onnx.ai"},{"key":"61_CR45","doi-asserted-by":"publisher","first-page":"13","DOI":"10.1007\/s41781-019-0027-2","volume":"3","author":"J Duarte","year":"2019","unstructured":"Duarte J et al (2019) FPGA-accelerated machine learning inference as a service for particle physics computing. Comput Softw Big Sci 3:13. https:\/\/doi.org\/10.1007\/s41781-019-0027-2","journal-title":"Comput Softw Big Sci"}],"container-title":["Computing and Software for Big Science"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/link.springer.com\/content\/pdf\/10.1007\/s41781-021-00061-3.pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/link.springer.com\/article\/10.1007\/s41781-021-00061-3\/fulltext.html","content-type":"text\/html","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/link.springer.com\/content\/pdf\/10.1007\/s41781-021-00061-3.pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2021,12,26]],"date-time":"2021-12-26T11:06:02Z","timestamp":1640516762000},"score":1,"resource":{"primary":{"URL":"https:\/\/link.springer.com\/10.1007\/s41781-021-00061-3"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2021,7,5]]},"references-count":45,"journal-issue":{"issue":"1","published-print":{"date-parts":[[2021,12]]}},"alternative-id":["61"],"URL":"https:\/\/doi.org\/10.1007\/s41781-021-00061-3","relation":{},"ISSN":["2510-2036","2510-2044"],"issn-type":[{"type":"print","value":"2510-2036"},{"type":"electronic","value":"2510-2044"}],"subject":[],"published":{"date-parts":[[2021,7,5]]},"assertion":[{"value":"26 July 2020","order":1,"name":"received","label":"Received","group":{"name":"ArticleHistory","label":"Article History"}},{"value":"8 June 2021","order":2,"name":"accepted","label":"Accepted","group":{"name":"ArticleHistory","label":"Article History"}},{"value":"5 July 2021","order":3,"name":"first_online","label":"First Online","group":{"name":"ArticleHistory","label":"Article History"}},{"order":1,"name":"Ethics","group":{"name":"EthicsHeading","label":"Declarations"}},{"value":"The authors declare that they have no conflict of interest.","order":2,"name":"Ethics","group":{"name":"EthicsHeading","label":"Conflict of interest"}},{"value":"The code is available at [, ] under MIT license.","order":3,"name":"Ethics","group":{"name":"EthicsHeading","label":"Code availability"}}],"article-number":"17"}}