Speaker: Dr Roger Melko, Department of Physics and astronomy, University Of Waterloo.
Title: Machine Learning the Many-Body Problem
Date: Thursday March 29, 2018
Time: 2:00 – 3:00 PM
Location: TRIUMF Auditorium, 4004 WesBrook Mall, Vancouver BC.
Abstract: Condensed matter physics is the study of the collective behaviour of infinitely complex assemblies of interacting electrons, magnetic moments, atoms or qubits. This complexity is reminiscent of the “curse of dimensionality” commonly encountered in machine learning. Despite this
curse, the machine learning community has developed techniques with remarkable abilities to classify, characterize and interpret complex sets of real-world data, such as images or natural languages. Here, we show that modern neural network architectures for supervised learning can be used to identify phases and phase transitions in a variety of condensed matter Hamiltonians. These neural networks can be trained to detect ordered
states, as well as topological states with no conventional order, directly from raw state configurations sampled theoretically or experimentally. Further, such configurations can be used to train a stochastic variant of a neural network, called a Restricted Boltzmann Machine (RBM), for use in unsupervised learning applications. We show how RBMs can be sampled much like a physical Hamiltonian to produce configurations useful for estimating physical observables. Finally, we examine the power of RBMs for the
efficient representation of classical and quantum Hamiltonians, and explore applications in quantum state tomography useful for near-term multi-quoit devices.
Biography: Dr. Melko is Canada Research Chair in Computational Many-Body Physics. Dr. Melko’s research interests involve strongly-correlated many-body systems, with a focus on emergent phenomena, ground state phases, phase transitions, quantum criticality, and entanglement. He emphasizes computational methods as a theoretical technique, in particular the development of state-of-the-art algorithms for the study of strongly-interacting systems. Dr. Melko’s work has employed Monte Carlo simulations and Density Matrix Renormalization Group methods to explore the low-temperature physics of classical and quantum magnetic materials, cold atoms in optical lattices, bosonic fluids and low-dimensional systems. He is particularly involved in studying microscopic models that display interesting quantum behavior in the bulk, such as superconducting, spin liquid, topological, superfluid or supersolid phases. He is also interested in broader ideas in computational physics, the development of efficient algorithms for simulating quantum mechanical systems on classical computers, and the relationship of these methods to the field of quantum information science.
Distinguished Lecture (IEEE NPSS)
Speaker: Dr Paule LeCoq, Senior Physicist at CERN, Geneva, Switzerland and Technical Director of uropean Center for Research in Medical Imaging in Marseilles
Title: The 10ps TOFPET challenge Myth or Reality?
Date: Monday May 7, 2018
Time: 2:00 – 3:00 PM
Location: TRIUMF Auditorium, 4004 WesBrook Mall, Vancouver BC.
Abstract: The future generation of radiation detectors is more and more demanding on timing performance for a wide range of applications, such as time of flight (TOF) techniques for PET cameras and particle identification in nuclear physics and high energy physics detectors, precise event time tagging in high luminosity accelerators and a number of photonic applications based on single photon detection. A target of 10ps coincidence time resolution in TOFPET scanners would introduce a paradigm shift in PET imaging. Besides resulting in on-line image formation, the localisation of annihilation events directly from their TOF provides ultimate use of the dose delivered to the patient to get the best Signal to Noise Ratio into the resulting image and offers a potential reduction of the scan duration and a direct access to the image during the scan itself. Reconstructionless TOF-PET also reduces efficiently undesired effects inherent to the PET detection, namely randoms and scatters when appropriately correlated to energy discrimination, hence contributing to reduce dose, scan duration and possibly scan cost while using very short-lived positron emitting isotopes.The time resolution of a scintillator-based detector is directly driven by the density of photoelectrons generated in the photodetector at the detection threshold. At the scintillator level it is related to the intrinsic light yield, the pulse shape (rise time and decay time) and the light transport from the gamma-ray conversion point to the photodetector. When aiming at 10ps time resolution fluctuations in the thermalization and relaxation time of hot electrons and holes generated by the interaction of ionization radiation with the crystal become important.This talk will review the different processes at work and evaluate if some of the transient phenomena taking place during the fast thermalization phase can be exploited to extract a time tag with a precision in the few ps range. Some considerations will also be given on the possibility to exploit quantum confinement for the production of ultrafast spontaneous or stimulated emission in semi-conductors.The light transport in the crystal is also an important source of time jitter. In particular light bouncing within the scintillator must be reduced as much as possible as it spreads the arrival time of photons on the photodetector and strongly reduces the light output by increasing the effect of light absorption within the crystal. A possible solution to overcome these problems is to improve the light extraction efficiency at the first hit of the photons on the crystal/photodetector coupling face by means of photonic crystals (PhCs) specifically designed to couple light propagation modes inside and outside the crystal at the limit of the total reflection angle.Finally the present limitations of the photodetectors, and in particular the SiPMs will be discussed and some R&D lines to meet the 10ps challenge will be presented.
Biography: Paul Lecoq has received his diploma as engineer in physics instrumentation at the Ecole Polytechnique de Grenoble in 1972, under the leadership of Nobel Laureate Louis Néel. After two years of work at the Nuclear Physics laboratory of the University of Montreal, Canada, he got his PhD in Nuclear Physics in 1974. Since then he has been working at CERN in 5 major international experiments on particle physics, two of them led by Nobel Laureates Samuel Ting and Carlo Rubbia. His action on detector instrumentation, and particularly on heavy inorganic scintillator materials has received a strong support from Georges Charpak. Member of a number of advisory committees and of international Societies he is since 2002 the promoter of the European Center for Research in Medical Imaging (Cerimed) presently being installed in Marseilles. He is an elected member of the European Academy of Sciences (2008).
Special Seminar
Distinguished Lecture (IEEE NPSS)