• Corpus ID: 226975720

Hierarchical clustering in particle physics through reinforcement learning

  title={Hierarchical clustering in particle physics through reinforcement learning},
  author={Johann Brehmer and Sebastian Macaluso and Duccio Pappadopulo and Kyle Cranmer},
Particle physics experiments often require the reconstruction of decay patterns through a hierarchical clustering of the observed final-state particles. We show that this task can be phrased as a Markov Decision Process and adapt reinforcement learning algorithms to solve it. In particular, we show that Monte-Carlo Tree Search guided by a neural policy can construct high-quality hierarchical clusterings and outperform established greedy and beam search baselines. 

Figures and Tables from this paper

A Living Review of Machine Learning for Particle Physics
This living review is a nearly comprehensive list of citations for those developing and applying deep learning approaches to experimental, phenomenological, or theoretical analyses, and will be updated as often as possible to incorporate the latest developments.
Exact and Approximate Hierarchical Clustering Using A
A new approach based on A* search is introduced that overcome the prohibitively large search space by combining A* with a novel trellis data structure and provides significantly improved theoretical bounds on the time and space complexity of A* for clustering.
Reframing Jet Physics with New Computational Methods
We reframe common tasks in jet physics in probabilistic terms, including jet reconstruction, Monte Carlo tuning, matrix element – parton shower matching for large jet multiplicity, and efficient


Jet grooming through reinforcement learning
It is shown that the grooming algorithm derived from the deep RL agent can match state-of-the-art techniques used at the Large Hadron Collider, resulting in improved mass resolution for boosted objects.
Compact Representation of Uncertainty in Hierarchical Clustering
This paper presents dynamic-programming algorithms and proofs for exact inference in hierarchical clustering that supports a wide range of hierarchical models and only requires a cluster compatibility function.
The frontier of simulation-based inference
The rapidly developing field of simulation-based inference is reviewed and the forces giving additional momentum to the field are identified to describe how the frontier is expanding.
Longitudinally-invariant k ⊥ -clustering algorithms for hadron-hadron collisions
Abstract We propose a version of the QCD-motivated “k⊥” jet-clustering algorithm for hadron-hadron collisions which is invariant under boosts along the beam directions. This leads to improved
Nondeterministic approach to tree-based jet substructure.
It is found that employing a set of trees substantially reduces the observed fluctuations in the pruned mass distribution, enhancing the likelihood of new particle discovery for a given integrated luminosity.
Better jet clustering algorithms
We investigate modifications to the k⊥-clustering jet algorithm which preserve the advantages of the original Durham algorithm while reducing non-perturbative corrections and providing better
Successive combination jet algorithm for hadron collisions.
  • Ellis, Soper
  • Physics, Medicine
    Physical review. D, Particles and fields
  • 1993
It is suggested that a successive combination style algorithm, similar to that used in [ital e][sup +][ital e[minus]] physics, might be useful also in hadron collisions, where cone style algorithms have been used previously.
General-purpose event generators for LHC physics
We review the physics basis, main features and use of general-purpose Monte Carlo event generators for the simulation of proton-proton collisions at the Large Hadron Collider. Topics included are:
The anti-$k_t$ jet clustering algorithm
The k_t and Cambridge/Aachen inclusive jet finding algorithms for hadron-hadron collisions can be seen as belonging to a broader class of sequential recombination jet algorithms, parametrised by the