Latent Space Refinement for Deep Generative Models

Deep generative models are becoming widely used across science and industry for a variety of purposes. A common challenge is achieving a precise implicit or explicit representation of the data probability density. Recent proposals have suggested using classifier weights to refine the learned density of deep generative models. We extend this idea to all types of generative models and show how latent space refinement via iterated generative modeling can circumvent topological obstructions and improve precision. This methodology also applies to cases were the target model is non-differentiable and has many internal latent dimensions which must be marginalized over before refinement. We demonstrate our Latent Space Refinement (LaSeR) protocol on a variety of examples, focusing on the combinations of Normalizing Flows and Generative Adversarial Networks.

[1]  T. Plehn,et al.  Understanding Event-Generation Networks via Uncertainties , 2021, SciPost Physics.

[2]  S. Carrazza,et al.  Compressing PDF sets using generative adversarial networks , 2021, The European Physical Journal C.

[3]  Sofia Vallecorsa,et al.  Validation of Deep Convolutional Generative Adversarial Networks for High Energy Physics Calorimeter Simulations , 2021, AAAI Spring Symposium: MLPS.

[4]  Vikram A. Saletore,et al.  Reduced Precision Strategies for Deep Learning: A High Energy Physics Generative Adversarial Network Use Case , 2021, ICPRAM.

[5]  Blaz Bortolato,et al.  Bump Hunting in Latent Space , 2021, Physical Review D.

[6]  D. Whiteson,et al.  Efficient sampling of constrained high-dimensional theoretical spaces with machine learning , 2021, The European Physical Journal C.

[7]  G. Kasieczka,et al.  Decoding Photons: Physics in the Latent Space of a BIB-AE Generative Network , 2021, EPJ Web of Conferences.

[8]  Suyong Choi,et al.  A data-driven event generator for Hadron Colliders using Wasserstein Generative Adversarial Network , 2021, Journal of the Korean Physical Society.

[9]  B. Nachman,et al.  Learning from many collider events at once , 2021, Physical Review D.

[10]  Stefan T. Radev,et al.  Measuring QCD Splittings with Invertible Networks , 2020, SciPost Physics.

[11]  A. Butter,et al.  How to GAN Event Unweighting , 2020, 2012.07873.

[12]  Y. Lai,et al.  Explainable machine learning of the underlying physics of high-energy particle collisions , 2020, Physics Letters B.

[13]  F. Ratnikov,et al.  Simulating the time projection chamber responses at the MPD detector using generative adversarial networks , 2020, The European Physical Journal C.

[14]  Javier Mauricio Duarte,et al.  Graph Generative Adversarial Networks for Sparse Data Generation in High Energy Physics , 2020, ArXiv.

[15]  Rob Verheyen,et al.  Phase space sampling and inference from weighted events with autoregressive flows , 2020, SciPost Physics.

[16]  Stefan Friedl.  Algebraic topology , 2020, Graduate Studies in Mathematics.

[17]  P. Baldi,et al.  SARM: Sparse Autoregressive Model for Scalable Generation of Sparse Images in Particle Physics , 2020, 2009.14017.

[18]  G. Kasieczka,et al.  DCTRGAN: improving the precision of generative models with reweighting , 2020, Journal of Instrumentation.

[19]  Kosei Dohi,et al.  Variational Autoencoders for Jet Simulation , 2020, 2009.04842.

[20]  Suyong Choi,et al.  Data-driven Estimation of Background Distribution through Neural Autoregressive Flows , 2020, 2008.03636.

[21]  P. Ambrozewicz,et al.  AI-based Monte Carlo event generator for electron-proton scattering , 2020 .

[22]  J. Arguin,et al.  Variational autoencoders for anomalous jet tagging , 2020, Physical Review D.

[23]  Kamil Deja,et al.  End-to-End Sinkhorn Autoencoder With Noise Generator , 2020, IEEE Access.

[24]  Ullrich Kothe,et al.  Invertible networks or partons to detector and back again , 2020, 2006.06685.

[25]  U. Seljak,et al.  Probabilistic Auto-Encoder , 2020, Trans. Mach. Learn. Res..

[26]  G. Kasieczka,et al.  Getting High: High Fidelity Simulation of High Granularity Calorimeters with High Speed , 2020, Computing and Software for Big Science.

[27]  Marcus A. Brubaker,et al.  Normalizing Flows: An Introduction and Review of Current Methods , 2020, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[28]  K. Cranmer,et al.  Flows for simultaneous manifold learning and density estimation , 2020, NeurIPS.

[29]  Danilo Jimenez Rezende,et al.  Equivariant flow-based sampling for lattice gauge theory , 2020, Physical review letters.

[30]  Yoshua Bengio,et al.  Your GAN is Secretly an Energy-based Model and You Should use Discriminator Driven Latent Sampling , 2020, NeurIPS.

[31]  Aaron C. Courville,et al.  Augmented Normalizing Flows: Bridging the Gap Between Generative Flows and Latent Variable Models , 2020, ArXiv.

[32]  H. Schulz,et al.  Event generation with normalizing flows , 2020, Physical Review D.

[33]  S. Schumann,et al.  Exploring phase space with Neural Importance Sampling , 2020, SciPost Physics.

[34]  Christina Gao,et al.  i- flow: High-dimensional integration and sampling with normalizing flows , 2020, Mach. Learn. Sci. Technol..

[35]  B. Nachman,et al.  Anomaly detection with density estimation , 2020, Physical Review D.

[36]  A. Butter,et al.  How to GAN event subtraction , 2019, SciPost Physics Core.

[37]  Wei Wei,et al.  Calorimetry with deep learning: particle simulation and reconstruction for collider physics , 2019, The European Physical Journal C.

[38]  Maurizio Pierini,et al.  Particle Generative Adversarial Networks for full-event simulation at the LHC and their application to pileup description , 2019, Journal of Physics: Conference Series.

[39]  Natalia Gimelshein,et al.  PyTorch: An Imperative Style, High-Performance Deep Learning Library , 2019, NeurIPS.

[40]  G. Kasieczka,et al.  How to GAN away Detector Effects , 2019, SciPost Physics.

[41]  C. Fanelli,et al.  DeepRICH: learning deeply Cherenkov detectors , 2019, Mach. Learn. Sci. Technol..

[42]  Patrick T. Komiske,et al.  OmniFold: A Method to Simultaneously Unfold All Observables. , 2019, Physical review letters.

[43]  Anthony L. Caterini,et al.  Relaxing Bijectivity Constraints with Continuously Indexed Normalising Flows , 2019, ICML.

[44]  SHiP Collaboration.  Fast simulation of muons produced at the SHiP experiment using Generative Adversarial Networks , 2019, Journal of Instrumentation.

[45]  S. Carrazza,et al.  Lund jet images from generative and cycle-consistent adversarial networks , 2019, The European Physical Journal C.

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

[47]  Tilman Plehn,et al.  How to GAN LHC events , 2019, SciPost Physics.

[48]  Sofia Vallecorsa,et al.  3D convolutional GAN for fast simulation , 2019, EPJ Web of Conferences.

[49]  W. Bhimji,et al.  Next Generation Generative Neural Networks for HEP , 2019, EPJ Web of Conferences.

[50]  Diederik P. Kingma,et al.  An Introduction to Variational Autoencoders , 2019, Found. Trends Mach. Learn..

[51]  M. S. Albergo,et al.  Flow-based generative models for Markov chain Monte Carlo in lattice field theory , 2019, Physical Review D.

[52]  D. Wipf,et al.  Diagnosing and Enhancing VAE Models , 2019, ICLR.

[53]  Sana Ketabchi Haghighat,et al.  DijetGAN: a Generative-Adversarial Network approach for the simulation of QCD dijet events at the LHC , 2019, Journal of High Energy Physics.

[54]  Benjamin Nachman,et al.  Machine learning templates for QCD factorization in the search for physics beyond the standard model , 2019, Journal of High Energy Physics.

[55]  Maurizio Pierini,et al.  LHC analysis-specific datasets with Generative Adversarial Networks , 2019, ArXiv.

[56]  Dmitry Ulyanov,et al.  Generative Models for Fast Calorimeter Simulation: the LHCb case> , 2018, EPJ Web of Conferences.

[57]  Saúl Alonso-Monsalve,et al.  Image-Based Model Parameter Optimization Using Model-Assisted Generative Adversarial Networks , 2018, IEEE Transactions on Neural Networks and Learning Systems.

[58]  Jan M. Pawlowski,et al.  Reducing autocorrelation times in lattice simulations with generative adversarial networks , 2018, Mach. Learn. Sci. Technol..

[59]  L. Pang,et al.  Regressive and generative neural networks for scalar field theory , 2018, Physical Review D.

[60]  Sven Krippendorf,et al.  GANs for generating EFT models , 2018, Physics Letters B.

[61]  S. Vallecorsa,et al.  Generative models for fast simulation , 2018, Journal of Physics: Conference Series.

[62]  P. Mendez Lorenzo,et al.  Three dimensional Generative Adversarial Networks for fast simulation , 2018, Journal of Physics: Conference Series.

[63]  David Rousseau,et al.  Further developments of FORM , 2018, Journal of Physics: Conference Series.

[64]  J. Monk,et al.  Deep learning as a parton shower , 2018, Journal of High Energy Physics.

[65]  Martin Erdmann,et al.  Precise Simulation of Electromagnetic Calorimeter Showers Using a Wasserstein Generative Adversarial Network , 2018, Computing and Software for Big Science.

[66]  Yisong Yue,et al.  Iterative Amortized Inference , 2018, ICML.

[67]  T. Trzciński,et al.  Generative Models for Fast Cluster Simulations in the TPC for the ALICE Experiment , 2018, Advances in Intelligent Systems and Computing.

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

[69]  Deepak Kar,et al.  Unfolding with Generative Adversarial Networks , 2018, 1806.00433.

[70]  Francesco Pandolfi,et al.  Fast and Accurate Simulation of Particle Detectors Using Generative Adversarial Networks , 2018, Computing and Software for Big Science.

[71]  Martin Erdmann,et al.  Generating and Refining Particle Detector Simulations Using the Wasserstein Distance in Adversarial Networks , 2018, Computing and Software for Big Science.

[72]  Alexander M. Rush,et al.  Semi-Amortized Variational Autoencoders , 2018, ICML.

[73]  Michela Paganini,et al.  CaloGAN: Simulating 3D High Energy Particle Showers in Multi-Layer Electromagnetic Calorimeters with Generative Adversarial Networks , 2017, ArXiv.

[74]  Michela Paganini,et al.  Controlling Physical Attributes in GAN-Accelerated Simulation of Electromagnetic Calorimeters , 2017, Journal of Physics: Conference Series.

[75]  Tom White,et al.  Generative Adversarial Networks: An Overview , 2017, IEEE Signal Processing Magazine.

[76]  Benjamin Nachman,et al.  Accelerating Science with Generative Adversarial Networks: An Application to 3D Particle Showers in Multilayer Calorimeters. , 2017, Physical review letters.

[77]  Luke de Oliveira,et al.  Learning Particle Physics by Example: Location-Aware Generative Adversarial Networks for Physics Synthesis , 2017, Computing and Software for Big Science.

[78]  Samy Bengio,et al.  Density estimation using Real NVP , 2016, ICLR.

[79]  R. Devon Hjelm,et al.  Iterative Refinement of the Approximate Posterior for Directed Belief Networks , 2015, NIPS.

[80]  Shakir Mohamed,et al.  Variational Inference with Normalizing Flows , 2015, ICML.

[81]  R. Zemel,et al.  Generative Moment Matching Networks , 2015, ICML.

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

[83]  Aaron C. Courville,et al.  Generative adversarial networks , 2014, Commun. ACM.

[84]  S. Mukherjee,et al.  The topology of probability distributions on manifolds , 2013, 1307.1123.

[85]  Masashi Sugiyama,et al.  Density Ratio Estimation in Machine Learning , 2012 .

[86]  L. Younes.  Shapes and Diffeomorphisms , 2010 .

[87]  Eric R. Ziegel,et al.  The Elements of Statistical Learning , 2003, Technometrics.

[88]  A. Kennedy,et al.  Hybrid Monte Carlo , 1988 .

[89]  D. Whiteson,et al.  Foundations of a Fast, Data-Driven, Machine-Learned Simulator , 2021 .

[90]  Nicolas Courty,et al.  POT: Python Optimal Transport , 2021, J. Mach. Learn. Res..

[91]  A. Hariri,et al.  Graph Generative Models for Fast Detector Simulations in High Energy Physics , 2021, ArXiv.

[92]  Michael I. Jordan,et al.  AUTO-ENCODING VARIATIONAL BAYES , 2020 .

[93]  Luke de Oliveira.  Tips and Tricks for Training GANs with Physics Constraints , 2017 .

[94]  Maurizio Pierini,et al.  Calorimetry with Deep Learning : Particle Classification , Energy Regression , and Simulation for High-Energy Physics , 2017 .