Parallel kt jet clustering algorithm

Abstract The numerical simulation allows to study the high energy particle physics. It plays important of role in the reconstruction and analyze of these experimental and theoretical researches. This requires a computer background with a large capacity. Jet physics is an intensively researched area, where the factorization process can be solved by algorithmic solutions. We studied parallelization of the kt cluster algorithms. This method allows to know the development of particles due to the collision of highenergy nucleus-nucleus. The Alice offline library contains the required modules to simulate the ALICE detector that is a dedicated Pb-Pb detector. Using this simulation we can generate input particles, that we can further analyzed by clustering them, reconstructing their jet structure. The FastJet toolkit is an efficient C++ implementation of the most widely used jet clustering algorithms, among them the kt clustering. Parallelizing the standard non-optimized version of this algorithm utilizing the available CPU architecture a 1:6 times faster runtime was achieved, paving the way to drastic performance increase using many-core architectures.

Download Full-text

Hierarchical clustering with deep Q-learning

Acta Universitatis Sapientiae Informatica ◽

10.2478/ausi-2018-0006 ◽

2018 ◽

Vol 10 (1) ◽

pp. 86-109

Author(s):

Richárd Forster ◽

Agnes Fülöp

Keyword(s):

Deep Learning ◽

Hierarchical Clustering ◽

Particle Physics ◽

Clustering Algorithms ◽

High Energy ◽

Jet Physics ◽

High Energy Particle ◽

Energy Particle ◽

Q Learning ◽

Underlying Graph

Abstract Following up on our previous study on applying hierarchical clustering algorithms to high energy particle physics, this paper explores the possibilities to use deep learning to generate models capable of processing the clusterization themselves. The technique chosen for training is reinforcement learning, that allows the system to evolve based on interactions between the model and the underlying graph. The result is a model, that by learning on a modest dataset of 10, 000 nodes during 70 epochs can reach 83, 77% precision for hierarchical and 86, 33% for high energy jet physics datasets in predicting the appropriate clusters.

Download Full-text

Hierarchical kt jet clustering for parallel architectures

Acta Universitatis Sapientiae Informatica ◽

10.1515/ausi-2017-0012 ◽

2017 ◽

Vol 9 (2) ◽

pp. 195-213

Author(s):

Richárd Forster ◽

Ágnes Fülöp

Keyword(s):

Hierarchical Clustering ◽

Particle Physics ◽

Clustering Algorithm ◽

Clustering Algorithms ◽

High Energy ◽

Theoretical Physics ◽

High Energy Particle ◽

Clustering Methods ◽

Hierarchical Clustering Methods ◽

Using Data

AbstractThe reconstruction and analyze of measured data play important role in the research of high energy particle physics. This leads to new results in both experimental and theoretical physics. This requires algorithm improvements and high computer capacity. Clustering algorithm makes it possible to get to know the jet structure more accurately. More granular parallelization of the kt cluster algorithms was explored by combining it with the hierarchical clustering methods used in network evaluations. The kt method allows to know the development of particles due to the collision of high-energy nucleus-nucleus. The hierarchical clustering algorithms works on graphs, so the particle information used by the standard kt algorithm was first transformed into an appropriate graph, representing the network of particles. Testing was done using data samples from the Alice offine library, which contains the required modules to simulate the ALICE detector that is a dedicated Pb-Pb detector. The proposed algorithm was compared to the FastJet toolkit's standard longitudinal invariant kt implementation. Parallelizing the standard non-optimized version of this algorithm utilizing the available CPU architecture proved to be 1:6 times faster, than the standard implementation, while the proposed solution in this paper was able to achieve a 12 times faster computing performance, also being scalable enough to efficiently run on GPUs.

Download Full-text