D. Bonacorsi

Luca Anzalone, Tommaso Diotalevi, Daniele Bonacorsi

Signal-background classification is a central problem in High-Energy Physics (HEP), that plays a major role for the discovery of new fundamental particles. A recent method -- the Parametric Neural Network (pNN) -- leverages multiple signal mass hypotheses as an additional input feature to effectively replace a whole set of individual classifiers, each providing (in principle) the best response for the corresponding mass hypothesis. In this work we aim at deepening the understanding of pNNs in light of real-world usage. We discovered several peculiarities of parametric networks, providing intuition, metrics, and guidelines to them. We further propose an alternative parametrization scheme, resulting in a new parametrized neural network architecture: the AffinePNN; along with many other generally applicable improvements, like the balanced training procedure. Finally, we extensively and empirically evaluate our models on the HEPMASS dataset, along its imbalanced version (called HEPMASS-IMB) we provide here for the first time, to further validate our approach. Provided results are in terms of the impact of the proposed design decisions, classification performance, and interpolation capability, as well.

Tommaso Diotalevi, Antonio Falabella, Barbara Martelli et al.

The distributed Grid infrastructure for High Energy Physics experiments at the Large Hadron Collider (LHC) in Geneva comprises a set of computing centres, spread all over the world, as part of the Worldwide LHC Computing Grid (WLCG). In Italy, the Tier-1 functionalities are served by the INFN-CNAF data center, which provides also computing and storage resources to more than twenty non-LHC experiments. For this reason, a high amount of logs are collected each day from various sources, which are highly heterogeneous and difficult to harmonize. In this contribution, a working implementation of a system that collects, parses and displays the log information from CNAF data sources and the investigation of a Machine Learning based predictive maintenance system, is presented.

Leticia Decker, Daniel Leite, Luca Giommi et al.

Detection of anomalous behaviors in data centers is crucial to predictive maintenance and data safety. With data centers, we mean any computer network that allows users to transmit and exchange data and information. In particular, we focus on the Tier-1 data center of the Italian Institute for Nuclear Physics (INFN), which supports the high-energy physics experiments at the Large Hadron Collider (LHC) in Geneva. The center provides resources and services needed for data processing, storage, analysis, and distribution. Log records in the data center is a stochastic and non-stationary phenomenon in nature. We propose a real-time approach to monitor and classify log records based on sliding time windows, and a time-varying evolving fuzzy-rule-based classification model. The most frequent log pattern according to a control chart is taken as the normal system status. We extract attributes from time windows to gradually develop and update an evolving Gaussian Fuzzy Classifier (eGFC) on the fly. The real-time anomaly monitoring system has to provide encouraging results in terms of accuracy, compactness, and real-time operation.

Leticia Decker, Daniel Leite, Fabio Viola et al.

Log-based predictive maintenance of computing centers is a main concern regarding the worldwide computing grid that supports the CERN (European Organization for Nuclear Research) physics experiments. A log, as event-oriented adhoc information, is quite often given as unstructured big data. Log data processing is a time-consuming computational task. The goal is to grab essential information from a continuously changeable grid environment to construct a classification model. Evolving granular classifiers are suited to learn from time-varying log streams and, therefore, perform online classification of the severity of anomalies. We formulated a 4-class online anomaly classification problem, and employed time windows between landmarks and two granular computing methods, namely, Fuzzy-set-Based evolving Modeling (FBeM) and evolving Granular Neural Network (eGNN), to model and monitor logging activity rate. The results of classification are of utmost importance for predictive maintenance because priority can be given to specific time intervals in which the classifier indicates the existence of high or medium severity anomalies.