Simulation assisted likelihood-free anomaly detection

Given the lack of evidence for new particle discoveries at the Large Hadron Collider (LHC), it is critical to broaden the search program. A variety of model-independent searches have been proposed, adding sensitivity to unexpected signals. There are generally two types of such searches: those that rely heavily on simulations and those that are entirely based on (unlabeled) data. This paper introduces a hybrid method that makes the best of both approaches. For potential signals that are resonant in one known feature, this new method first learns a parameterized reweighting function to morph a given simulation to match the data in sidebands. This function is then interpolated into the signal region and then the reweighted background-only simulation can be used for supervised learning as well as for background estimation. The background estimation from the reweighted simulation allows for non-trivial correlations between features used for classification and the resonant feature. A dijet search with jet substructure is used to illustrate the new method. Future applications of Simulation Assisted Likelihood-free Anomaly Detection (SALAD) include a variety of final states and potential combinations with other model-independent approaches.

[1]  J. Caudron,et al.  A strategy for a general search for new phenomena using data-derived signal regions and its application within the ATLAS experiment , 2018, 1807.07447.

[2]  M. White,et al.  Does SUSY have friends? A new approach for LHC event analysis , 2019, Journal of High Energy Physics.

[3]  D. Whiteson,et al.  Deep Learning and Its Application to LHC Physics , 2018, Annual Review of Nuclear and Particle Science.

[4]  Pierre Baldi,et al.  Parameterized neural networks for high-energy physics , 2016, The European Physical Journal C.

[5]  J. G. Contreras,et al.  A General Search for New Phenomena at HERA , 2007 .

[6]  Anders Andreassen,et al.  OmniFold: A Method to Simultaneously Unfold All Observables. , 2020, Physical review letters.

[7]  J. P. Fernández,et al.  Model-Independent Global Search for New High-pT Physics at CDF , 2007 .

[8]  D. Whiteson,et al.  Model-independent and quasi-model-independent search for new physics at CDF , 2008 .

[9]  B. Nachman.  A guide for deploying Deep Learning in LHC searches: How to achieve optimality and account for uncertainty , 2019, SciPost Physics.

[10]  A General search for new phenomena in ep scattering at HERA , 2004 .

[11]  M. Cacciari,et al.  The anti-$k_t$ jet clustering algorithm , 2008, 0802.1189.

[12]  Y. Arnoud,et al.  A QUASI MODEL INDEPENDENT SEARCH FOR NEW HIGH PT PHYSICS AT D0 , 2002 .

[13]  G. Menardi,et al.  Nonparametric semi-supervised classification with application to signal detection in high energy physics , 2021, Statistical Methods & Applications.

[14]  D. Shih,et al.  Searching for new physics with deep autoencoders , 2018, Physical Review D.

[15]  Jun Zhao,et al.  Supervised Deep Learning in High Energy Phenomenology: a Mini Review , 2019, Communications in Theoretical Physics.

[16]  M. Gigg,et al.  Herwig++ physics and manual , 2008, 0803.0883.

[17]  M. Cacciari,et al.  Dispelling the N3 myth for the kt jet-finder , 2005, hep-ph/0512210.

[18]  Stefan Wunsch,et al.  Reducing the Dependence of the Neural Network Function to Systematic Uncertainties in the Input Space , 2019, Computing and Software for Big Science.

[19]  J. Thaler,et al.  Identifying boosted objects with N-subjettiness , 2010, 1011.2268.

[20]  Philip Harris,et al.  Machine learning uncertainties with adversarial neural networks , 2018, The European Physical Journal C.

[21]  Maria Spiropulu,et al.  Variational autoencoders for new physics mining at the Large Hadron Collider , 2018, Journal of High Energy Physics.

[22]  G. Kasieczka,et al.  DisCo Fever: Robust Networks Through Distance Correlation , 2020, 2001.05310.

[23]  J. S. Hoftun,et al.  Quasi-model-independent search for new high p(T) physics at D0. , 2000, Physical review letters.

[24]  T. Roy,et al.  A robust anomaly finder based on autoencoder , 2019, 1903.02032.

[25]  B. Nachman,et al.  Jet substructure at the Large Hadron Collider: A review of recent advances in theory and machine learning , 2017, Physics Reports.

[26]  Gilles Louppe,et al.  Constraining Effective Field Theories with Machine Learning. , 2018, Physical review letters.

[27]  B. Nachman,et al.  Extending the search for new resonances with machine learning , 2019, Physical Review D.

[28]  A. Mertens.  New features in Delphes 3 , 2015 .

[29]  Tsuyoshi Murata,et al.  {m , 1934, ACML.

[30]  D. Whiteson,et al.  The Unexplored Landscape of Two-body Resonances , 2016, Acta Physica Polonica B.

[31]  E. S. Pearson,et al.  On the Problem of the Most Efficient Tests of Statistical Hypotheses , 1933 .

[32]  Benjamin Nachman,et al.  AI Safety for High Energy Physics , 2019, ArXiv.

[33]  J. Favereau,et al.  DELPHES 3: a modular framework for fast simulation of a generic collider experiment , 2013, Journal of High Energy Physics.

[34]  H. Collaboration,et al.  A General Search for New Phenomena at HERA , 2009, 0901.0507.

[35]  B. Nachman,et al.  Anomaly Detection for Resonant New Physics with Machine Learning. , 2018, Physical review letters.

[36]  Gilles Louppe,et al.  Mining gold from implicit models to improve likelihood-free inference , 2018, Proceedings of the National Academy of Sciences.

[37]  B. Nachman,et al.  Neural networks for full phase-space reweighting and parameter tuning , 2019, Physical Review D.

[38]  F. Cardillo.  A general search for new phenomena with the ATLAS detector in pp collisions at $\sqrt{s}$ = 8 TeV , 2014 .

[39]  Gilles Louppe,et al.  Learning to Pivot with Adversarial Networks , 2016, NIPS.

[40]  Jimmy Ba,et al.  Adam: A Method for Stochastic Optimization , 2014, ICLR.

[41]  S. Mrenna,et al.  Pythia 6.3 physics and manual , 2003, hep-ph/0308153.

[42]  E. al.,et al.  Global search for new physics with 2.0 fb(-1) at CDF , 2008, 0809.3781.

[43]  P. Harris,et al.  Thinking outside the ROCs: Designing Decorrelated Taggers (DDT) for jet substructure , 2016, 1603.00027.

[44]  Gilles Louppe,et al.  Approximating Likelihood Ratios with Calibrated Discriminative Classifiers , 2015, 1506.02169.

[45]  Patrick T. Komiske,et al.  Energy flow networks: deep sets for particle jets , 2018, Journal of High Energy Physics.

[46]  M. Cacciari,et al.  FastJet user manual , 2011, 1111.6097.

[47]  Mike Williams,et al.  uBoost: a boosting method for producing uniform selection efficiencies from multivariate classifiers , 2013, 1305.7248.

[48]  Garry Tamlyn,et al.  Music , 1993 .

[49]  Layne Bradshaw,et al.  Mass Agnostic Jet Taggers , 2019 .

[50]  R. D’Agnolo,et al.  Learning new physics from a machine , 2018, Physical Review D.

[51]  Yuan Yu,et al.  TensorFlow: A system for large-scale machine learning , 2016, OSDI.

[52]  Gregor Kasieczka,et al.  QCD or what? , 2018, SciPost Physics.

[53]  J. Sculli,et al.  Search for new physics in aμX data at DØ using SLEUTH: A quasi-model-independent search strategy for new physics , 2000 .

[54]  J. A. Aguilar-Saavedra,et al.  A generic anti-QCD jet tagger , 2017, 1709.01087.

[55]  J. Thaler,et al.  Maximizing boosted top identification by minimizing N-subjettiness , 2011, 1108.2701.

[56]  A. Simone,et al.  Guiding new physics searches with unsupervised learning , 2018, The European Physical Journal C.

[57]  M. Spannowsky,et al.  Adversarially-trained autoencoders for robust unsupervised new physics searches , 2019, Journal of High Energy Physics.

[58]  Peter Skands,et al.  A brief introduction to PYTHIA 8.1 , 2007, Comput. Phys. Commun..

[59]  Tao Liu,et al.  Novelty Detection Meets Collider Physics , 2018, Physical Review D.

[60]  Jernej F. Kamenik,et al.  Uncovering latent jet substructure , 2019, Physical Review D.

[61]  L. Xia.  QBDT, a new boosting decision tree method with systematical uncertainties into training for High Energy Physics , 2018, Nuclear Instruments and Methods in Physics Research Section A: Accelerators, Spectrometers, Detectors and Associated Equipment.

[62]  Vm Joint Nucl Res Inst Dubna Russia. Abazov,et al.  Quasi-model-independent search for new physics at large transverse momentum , 2001 .

[63]  Gilles Louppe,et al.  A guide to constraining effective field theories with machine learning , 2018, Physical Review D.

[64]  Kazuhiro Terao,et al.  Machine learning at the energy and intensity frontiers of particle physics , 2018, Nature.

[65]  J. Favereau,et al.  DELPHES 3: A modular framework for fast-simulation of generic collider experiments , 2014 .

[66]  D. Whiteson,et al.  The motivation and status of two-body resonance decays after the LHC Run 2 and beyond , 2019, Journal of High Energy Physics.

[67]  B. Nachman,et al.  Convolved substructure: analytically decorrelating jet substructure observables , 2017, 1710.06859.

[68]  Pierre Baldi,et al.  Decorrelated jet substructure tagging using adversarial neural networks , 2017, Physical Review D.

[69]  M. Pierini,et al.  Learning multivariate new physics , 2019, The European Physical Journal C.