arXiv subjects

Georgios Roumpos

Publications and source records attributed to Georgios Roumpos.

10 recordsLinked to original sources

arXiv2020

Modeling Information Need of Users in Search Sessions

Users issue queries to Search Engines, and try to find the desired information in the results produced. They repeat this process if their information need is not met at the first place. It is crucial to identify the important words in a query that depict the actual information need of the user and will determine the course of a search session. To this end, we propose a sequence-to-sequence based neural architecture that leverages the set of past queries issued by users, and results that were explored by them. Firstly, we employ our model for predicting the words in the current query that are important and would be retained in the next query. Additionally, as a downstream application of our model, we evaluate it on the widely popular task of next query suggestion. We show that our intuitive strategy of capturing information need can yield superior performance at these tasks on two large real-world search log datasets.

cs.IR↗

arXiv2019

High Performance Monte Carlo Simulation of Ising Model on TPU Clusters

Large-scale deep learning benefits from an emerging class of AI accelerators. Some of these accelerators' designs are general enough for compute-intensive applications beyond AI and Cloud TPU is one such example. In this paper, we demonstrate a novel approach using TensorFlow on Cloud TPU to simulate the two-dimensional Ising Model. TensorFlow and Cloud TPU framework enable the simple and readable code to express the complicated distributed algorithm without compromising the performance. Our code implementation fits into a small Jupyter Notebook and fully utilizes Cloud TPU's efficient matrix operation and dedicated high speed inter-chip connection. The performance is highly competitive: it outperforms the best published benchmarks to our knowledge by 60% in single-core and 250% in multi-core with good linear scaling. When compared to Tesla V100 GPU, the single-core performance maintains a ~10% gain. We also demonstrate that using low precision arithmetic---bfloat16---does not compromise the correctness of the simulation results.

cs.DC↗

arXiv2017

TensorFlow Estimators: Managing Simplicity vs. Flexibility in High-Level Machine Learning Frameworks

We present a framework for specifying, training, evaluating, and deploying machine learning models. Our focus is on simplifying cutting edge machine learning for practitioners in order to bring such technologies into production. Recognizing the fast evolution of the field of deep learning, we make no attempt to capture the design space of all possible model architectures in a domain- specific language (DSL) or similar configuration language. We allow users to write code to define their models, but provide abstractions that guide develop- ers to write models in ways conducive to productionization. We also provide a unifying Estimator interface, making it possible to write downstream infrastructure (e.g. distributed training, hyperparameter tuning) independent of the model implementation. We balance the competing demands for flexibility and simplicity by offering APIs at different levels of abstraction, making common model architectures available out of the box, while providing a library of utilities designed to speed up experimentation with model architectures. To make out of the box models flexible and usable across a wide range of problems, these canned Estimators are parameterized not only over traditional hyperparameters, but also using feature columns, a declarative specification describing how to interpret input data. We discuss our experience in using this framework in re- search and production environments, and show the impact on code health, maintainability, and development speed.

cs.DC↗

arXiv2015

Spatial correlation of two-dimensional Bosonic multi-mode condensates

We studied the spatial coherence of a Bosonic two-dimensional multi-mode condensate both through measurements and simulations. It is shown that condensates with a constant spatial density must be described as the superposition of several quantized modes which reduces the overall coherence. In this case, the spatial coherence can appear to decay faster than allowed by the Berezinskii-Kosterlitz-Thouless (BKT) theory. However, we find through spectroscopic measurements that the individual modes show a slower decay of the spatial coherence than the overall system.

cond-mat.quant-gas↗

arXiv2014

Algebraic order and the Berezinskii-Kosterlitz-Thouless transition in an exciton-polariton gas

We observe quasi-long range coherence in a two-dimensional condensate of exciton-polaritons. Our measurements are the first to confirm that the spatial correlation algebraically decays with a slow power-law, whose exponent quantitatively behaves as predicted by the Berezinskii-Kosterlitz-Thouless theory. The exciton-polaritons are created by non-resonant optical pumping of a micro-cavity sample with embedded GaAs quantum-wells at liquid helium temperature. Michelson interference is used to measure the coherence of the photons emitted by decaying exciton-polaritons.

cond-mat.quant-gas↗

arXiv2010

Single vortex-antivortex pair in an exciton polariton condensate

In a homogeneous two-dimensional system at non-zero temperature, although there can be no ordering of infinite range, a superfluid phase is predicted for a Bose liquid. The stabilization of phase in this superfluid regime is achieved by the formation of bound vortex-antivortex pairs. It is believed that several different systems share this common behaviour, when the parameter describing their ordered state has two degrees of freedom, and the theory has been tested for some of them. However, there has been no direct experimental observation of the phase stabilization mechanism by a bound pair. Here we present an experimental technique that can identify a single vortex-antivortex pair in a two-dimensional exciton polariton condensate. The pair is generated by the inhomogeneous pumping spot profile, and is revealed in the time-integrated phase maps acquired using Michelson interferometry, which show that the condensate phase is only locally disturbed. Numerical modelling based on open dissipative Gross-Pitaevskii equation suggests that the pair evolution is quite different in this non-equilibrium system compared to atomic condensates. Our results demonstrate that the exciton polariton condensate is a unique system for studying two-dimensional superfluidity in a previously inaccessible regime.

cond-mat.quant-gas↗

arXiv2010

Gain-induced trapping of microcavity exciton polariton condensates

We have performed real and momentum space spectroscopy of exciton polariton condensates in a GaAs-based microcavity under non-resonant excitation with an intensity stabilized laser. An effective trapping mechanism is revealed, which is due to the stimulated scattering gain inside the finite excitation spot combined with the short lifetime. We observe several quantized modes while the lowest state shows Heisenberg-limited real and momentum space distributions. The experimental findings are qualitatively reproduced by an open dissipative Gross-Pitaevskii equation model.

cond-mat.quant-gas↗

arXiv2009

Signature of the microcavity exciton-polariton relaxation mechanism in the polarization of emitted light

We have performed real and momentum space spin-dependent spectroscopy of spontaneously formed exciton polariton condensates for a non-resonant pumping scheme. Under linearly polarized pump, our results can be understood in terms of spin-dependent Boltzmann equations in a two-state model. This suggests that relaxation into the ground state occurs after multiple phonon scattering events and only one polariton-polariton scattering. For the circular pumping case, in which only excitons of one spin are injected, a bottleneck effect is observed, implying inefficient relaxation.

cond-mat.mes-hall↗

arXiv2008

GaAs Microcavity Exciton-Polaritons in a Trap

We present a simple method to create an in-plane lateral potential in a semiconductor microcavity using a metal thin-film. Two types of potential are produced: a circular aperture and a one-dimensional (1D) periodic grating pattern. The amplitude of the potential induced by a 24 nm-6 nm Au/Ti film is on the order of a few hundreds of ueV measured at 6 ~ 8 K. Since the metal layer makes the electromagnetic fields to be close to zero at the metal-semiconductor interface, the photon mode is confined more inside of the cavity. As a consequence, the effective cavity length is reduced under the metal film, and the corresponding cavity resonance is blue-shifted. Our experimental results are in a good agreement with theoretical estimates. In addition, by applying a DC electric voltage to the metal film, we are able to modify the quantum well exciton mode due to the quantum confined Stark effect, inducing a ~ 1 meV potential at ~ 20 kV/cm. Our method produces a controllable in-plane spatial trap potential for lower exciton-polaritons (LPs), which can be a building block towards 1D arrays and 2D lattices of LP condensates.

cond-mat.mes-hall↗

arXiv2007

Quantum simulation of spin ordering with nuclear spins in a solid state lattice

An experiment demonstrating the quantum simulation of a spin-lattice Hamiltonian is proposed. Dipolar interactions between nuclear spins in a solid state lattice can be modulated by rapid radio-frequency pulses. In this way, the effective Hamiltonian of the system can be brought to the form of an antiferromagnetic Heisenberg model with long range interactions. Using a semiconducting material with strong optical properties such as InP, cooling of nuclear spins could be achieved by means of optical pumping. An additional cooling stage is provided by adiabatic demagnetization in the rotating frame (ADRF) down to a nuclear spin temperature at which we expect a phase transition from a paramagnetic to antiferromagnetic phase. This phase transition could be observed by probing the magnetic susceptibility of the spin-lattice. Our calculations suggest that employing current optical pumping technology, observation of this phase transition is within experimental reach.

quant-ph↗