http://webgamma.ific.uv.es/gamma/
Who are we? What do we do? Where?
The science we do:
• Structure of atomic nuclei and reactions
• Nuclear astrophysics
• NS in HS of SE / n_TOF
• NS in EE of SE/ RIBs The technology we do:
• Nuclear instrumentation
• Data for reactor safety
• Radioactivity monitoring apps
ISOLDE/CERN, nTOF/CERN, GSI/FAIR, JYFL, RIKEN, GANIL, RCNP, LNLegnaro,
…
EAR2 20m EAR1 185 m Funding ID:
-FPA-2017 (National) -ERC-CoG (European)
Interested in COMCHA:
Victor Babiano (PhD-Student), Ion Ladarescu
(Eng.), Luis Caballero (Postdoc S8A), Kike
Nacher (CT), Jose Luis Taín (IC), Alejandro
Algora (CT), César Domingo (CT)
V.Babiano et al. “γ-Ray position reconstruction in large monolithic LaCl3(Ce) crystals with SiPM readout”, submitted to NIM-A, https://arxiv.org/abs/1811.05469
Neural Networks @ HYMNS Project / Victor Babiano, Ion Ladarescu, Luis Caballero & César Domingo
• 3D g-ray hit position reconstruction in large monolithic scintillation crystals with SiPM readout
• CERN ROOT TMultiPerceptron Classes
• BFGS Learning Method
• Typical NN-structure 64iN/64aN/2oN
Our challenges:
• Improve NN-training time (right now about 48h for 35x35 points/1-3MEvents/point)
• Attempt more complex NNs: 256ANs
Our next steps/objectives:
• Parallelize our code /either CUDA-GPUs or Multithread
• NN-based g-ray imaging algorithms (Compton)
Bayesian Network @ TAS Project / Kike Nácher, Alejandro Algora & Jose Luis Taín
- State-of-the-art for TAS analysis is based on a BN where the JPD is the decay-response function of the detector.
This is solved using a ML-EM in a constrained phase-space of the problem.
Challenges and objectives:
- Expand the BN in the observational domain by adding g-ray intensities and multiplicities, and in the parameters- domain including branching ratios.
- Explore novel approaches, instead of the MLEM-approach use Markov-Chain MC and/or Genetic Algorithms.
Further interests & ideas requiring Computing Resources:
- Use of ANNs for background-reduction using pulse-shape analysis in neutron-detectors at underground facilities (Canfranc, etc).
- Use of parallelized CP for several nuclear physics calculations:
* Neutron interaction simulations with Geant4-MultiThread (n_TOF, neutron-sensitivity backgrounds analysis)
* Scintillation photons in large monolithic crystals (high CPU power) for developing position-reconstruction algorithms (HYMNS).
Geant4 Multithread MC code already developed, just need more CPU power to run our new studies.
P.Olleros et al., “On the performance of large monolithic LaCl3(Ce) crystals coupled to pixelated silicon photosensors”, JINST 13 (2018)
