-
Ai-Sampler: Adversarial Learning of Markov kernels with involutive maps
Authors:
Evgenii Egorov,
Ricardo Valperga,
Efstratios Gavves
Abstract:
Markov chain Monte Carlo methods have become popular in statistics as versatile techniques to sample from complicated probability distributions. In this work, we propose a method to parameterize and train transition kernels of Markov chains to achieve efficient sampling and good mixing. This training procedure minimizes the total variation distance between the stationary distribution of the chain…
▽ More
Markov chain Monte Carlo methods have become popular in statistics as versatile techniques to sample from complicated probability distributions. In this work, we propose a method to parameterize and train transition kernels of Markov chains to achieve efficient sampling and good mixing. This training procedure minimizes the total variation distance between the stationary distribution of the chain and the empirical distribution of the data. Our approach leverages involutive Metropolis-Hastings kernels constructed from reversible neural networks that ensure detailed balance by construction. We find that reversibility also implies $C_2$-equivariance of the discriminator function which can be used to restrict its function space.
△ Less
Submitted 4 June, 2024;
originally announced June 2024.
-
Nature of M31 gamma-ray halo in relation to dark matter annihilation
Authors:
Andrei E. Egorov
Abstract:
The present work analyzes various aspects of M31 gamma-ray halo emission in its relation to annihilating dark matter (DM). The main aspect is the predicted effect of asymmetry of the intensity of emission due to inverse Compton scattering (ICS) of a possible population of relativistic electrons and positrons ($e^\pm$) in the galactic halo on starlight photons. This asymmetry is expected to exist a…
▽ More
The present work analyzes various aspects of M31 gamma-ray halo emission in its relation to annihilating dark matter (DM). The main aspect is the predicted effect of asymmetry of the intensity of emission due to inverse Compton scattering (ICS) of a possible population of relativistic electrons and positrons ($e^\pm$) in the galactic halo on starlight photons. This asymmetry is expected to exist around the major galactic axis, and arises due to anisotropy of the interstellar radiation field and the inclination of M31. ICS emission and its asymmetry were modeled by GALPROP code for the trial case of $e^\pm$ generated by annihilating weakly interacting massive particles (WIMPs) with various properties. The asymmetry was obtained to appear at photon energies above $\sim$ 0.1 MeV. Morphological and spectral properties of the asymmetry were studied in detail. Potential observational detection of the asymmetry may allow to infer the leptonic fraction in the emission generation mechanism, thus providing valuable inferences for understanding the nature of M31 gamma-ray halo emission. Specific asymmetry predictions were made for the recently claimed DM interpretation of the outer halo emission. The paper also studied the role of secondary -- ICS and bremsstrahlung -- emissions due to DM annihilation for that interpretation. And, finally, the latter was shown to be somewhat restricted by the recently derived WIMP constraints from radio data on M31.
△ Less
Submitted 5 August, 2023; v1 submitted 12 May, 2023;
originally announced May 2023.
-
The END: An Equivariant Neural Decoder for Quantum Error Correction
Authors:
Evgenii Egorov,
Roberto Bondesan,
Max Welling
Abstract:
Quantum error correction is a critical component for scaling up quantum computing. Given a quantum code, an optimal decoder maps the measured code violations to the most likely error that occurred, but its cost scales exponentially with the system size. Neural network decoders are an appealing solution since they can learn from data an efficient approximation to such a map** and can automaticall…
▽ More
Quantum error correction is a critical component for scaling up quantum computing. Given a quantum code, an optimal decoder maps the measured code violations to the most likely error that occurred, but its cost scales exponentially with the system size. Neural network decoders are an appealing solution since they can learn from data an efficient approximation to such a map** and can automatically adapt to the noise distribution. In this work, we introduce a data efficient neural decoder that exploits the symmetries of the problem. We characterize the symmetries of the optimal decoder for the toric code and propose a novel equivariant architecture that achieves state of the art accuracy compared to previous neural decoders.
△ Less
Submitted 14 April, 2023;
originally announced April 2023.
-
Updated constraints on WIMP dark matter by radio observations of M31 -- all annihilation channels
Authors:
Andrei E. Egorov
Abstract:
The present work derived the robust constraints on annihilating WIMP parameters utilizing new radio observations of M31, as well as new studies of its DM distribution and other properties. The characteristics of emission due to DM annihilation were computed in the frame of 2D galactic model employing GALPROP code adapted specifically for M31. This enabled us to refine various inaccuracies of previ…
▽ More
The present work derived the robust constraints on annihilating WIMP parameters utilizing new radio observations of M31, as well as new studies of its DM distribution and other properties. The characteristics of emission due to DM annihilation were computed in the frame of 2D galactic model employing GALPROP code adapted specifically for M31. This enabled us to refine various inaccuracies of previous studies on the subject. DM constraints were obtained for all possible annihilation channels except $χχ\rightarrow γγ,ν\barν$. A wide variety of radio data was utilized in the frequency range $\approx$(0.1--10) GHz. As the result the thermal WIMP lighter than fiducially $\approx$(40--70) GeV was excluded in the case of light primary annihilation products $χχ\rightarrow τ^+τ^-,μ^+μ^-,gg,c\overline{c},u\overline{u},d\overline{d},s\overline{s},e^+e^-,b\overline{b}$ (and an arbitrary combination of them). Heavier WIMP, which can annihilate to $W^+W^-,Z^0Z^0,t\overline{t},hh$; can not be probed at the level of thermal cross section, unless one assumes the optimistic cases of DM density and magnetic field distributions in M31. In summary, $m_x \gtrsim 40$ GeV represents the fiducial channel-independent mass limit for the thermal WIMP with the full uncertainty range estimated to be $\approx$(20--90) GeV. The obtained exclusions are competitive to those from Fermi-LAT observations of dwarfs and AMS-02 measurements of antiprotons. Our constraints significantly restrict the opportunity to explain the gamma-ray outer halo of M31 by annihilating DM. And, finally, we questioned the possibility claimed in other studies to robustly constrain heavy thermal WIMP with $m_x > 100$ GeV by radio data on M31.
△ Less
Submitted 20 February, 2023; v1 submitted 30 August, 2022;
originally announced August 2022.
-
Updated constraints on WIMP dark matter annihilation by radio observations of M31
Authors:
Andrei E. Egorov
Abstract:
The present work derived the robust constraints on annihilating WIMP parameters utilizing new radio observations of M31, as well as new studies of its dark matter distribution and other properties. The characteristics of emission due to DM annihilation were computed in the frame of 2D galactic model employing GALPROP code adapted specifically for M31. This enabled us to refine various inaccuracies…
▽ More
The present work derived the robust constraints on annihilating WIMP parameters utilizing new radio observations of M31, as well as new studies of its dark matter distribution and other properties. The characteristics of emission due to DM annihilation were computed in the frame of 2D galactic model employing GALPROP code adapted specifically for M31. This enabled us to refine various inaccuracies of previous studies on the subject. DM constraints were obtained for two representative annihilation channels: $χχ\rightarrow b\overline{b}$ and $χχ\rightarrow τ^+τ^-$. A wide variety of radio data was utilized in the frequency range $\approx$(0.1--10) GHz. As the result the thermal WIMP lighter than fiducially $\approx$ 70 GeV for $b\overline{b}$ channel and $\approx$ 40 GeV for $τ^+τ^-$ was excluded. The corresponding mass threshold uncertainty ranges were estimated to be 20--210 GeV and 18--89 GeV. The obtained exclusions are competitive to those from Fermi-LAT observations of dwarfs and AMS-02 measurements of antiprotons. Our constraints do not exclude the explanation of the gamma-ray outer halo of M31 and the Galactic center excess by annihilating DM. The thermal WIMP with $m_x \approx 70$ GeV, which explains the outer halo, would make a significant contribution to the non-thermal radio flux in M31 nucleus, fitting well both the spectrum and morphology. And, finally, we questioned the possibility claimed in other studies to robustly constrain heavy thermal WIMP with $m_x > 100$ GeV by radio data on M31.
△ Less
Submitted 7 July, 2022; v1 submitted 2 May, 2022;
originally announced May 2022.
-
LRWR: Large-Scale Benchmark for Lip Reading in Russian language
Authors:
Evgeniy Egorov,
Vasily Kostyumov,
Mikhail Konyk,
Sergey Kolesnikov
Abstract:
Lipreading, also known as visual speech recognition, aims to identify the speech content from videos by analyzing the visual deformations of lips and nearby areas. One of the significant obstacles for research in this field is the lack of proper datasets for a wide variety of languages: so far, these methods have been focused only on English or Chinese. In this paper, we introduce a naturally dist…
▽ More
Lipreading, also known as visual speech recognition, aims to identify the speech content from videos by analyzing the visual deformations of lips and nearby areas. One of the significant obstacles for research in this field is the lack of proper datasets for a wide variety of languages: so far, these methods have been focused only on English or Chinese. In this paper, we introduce a naturally distributed large-scale benchmark for lipreading in Russian language, named LRWR, which contains 235 classes and 135 speakers. We provide a detailed description of the dataset collection pipeline and dataset statistics. We also present a comprehensive comparison of the current popular lipreading methods on LRWR and conduct a detailed analysis of their performance. The results demonstrate the differences between the benchmarked languages and provide several promising directions for lipreading models finetuning. Thanks to our findings, we also achieved new state-of-the-art results on the LRW benchmark.
△ Less
Submitted 14 September, 2021;
originally announced September 2021.
-
Gamma- and Cosmic-Ray Observations with the GAMMA-400 Gamma-Ray Telescope
Authors:
N. P. Topchiev,
A. M. Galper,
I. V. Arkhangelskaja,
A. I. Arkhangelskiy,
A. V. Bakaldin,
R. A. Cherniy,
I. V. Chernysheva,
E. N. Gudkova,
Yu. V. Gusakov,
O. D. Dalkarov,
A. E. Egorov,
M. D. Kheymits,
M. G. Korotkov,
A. A. Leonov,
A. G. Malinin,
V. V. Mikhailov,
A. V. Mikhailova,
P. Yu. Minaev,
N. Yu. Pappe,
M. V. Razumeyko,
M. F. Runtso,
Yu. I. Stozhkov,
S. I. Suchkov,
Yu. T. Yurkin
Abstract:
The future space-based GAMMA-400 gamma-ray telescope will operate onboard the Russian astrophysical observatory in a highly elliptic orbit during 7 years to observe Galactic plane, Galactic Center, Fermi Bubbles, Crab, Vela, Cygnus X, Geminga, Sun, and other regions and measure gamma- and cosmic-ray fluxes. Observations will be performed in the point-source mode continuously for a long time (~100…
▽ More
The future space-based GAMMA-400 gamma-ray telescope will operate onboard the Russian astrophysical observatory in a highly elliptic orbit during 7 years to observe Galactic plane, Galactic Center, Fermi Bubbles, Crab, Vela, Cygnus X, Geminga, Sun, and other regions and measure gamma- and cosmic-ray fluxes. Observations will be performed in the point-source mode continuously for a long time (~100 days). GAMMA-400 will measure gamma rays in the energy range from ~20 MeV to several TeV and cosmic-ray electrons + positrons up to several tens TeV. GAMMA-400 instrument will have very good angle and energy resolutions, high separation efficiency of gamma rays from cosmic-ray background, as well as electrons + positrons from protons. The main feature of GAMMA-400 is the unprecedented angular resolution for energies >30 GeV better than the space-based and ground-based gamma-ray telescopes by a factor of 5-10. GAMMA-400 observations will permit to resolve gamma rays from annihilation or decay of dark matter particles, identify many discrete sources, clarify the structure of extended sources, specify the data on cosmic-ray electron + positron spectra.
△ Less
Submitted 21 January, 2022; v1 submitted 28 August, 2021;
originally announced August 2021.
-
On a possible origin of the gamma-ray excess around the Galactic Center
Authors:
Dmitry O. Chernyshov,
Andrei E. Egorov,
Vladimir A. Dogiel,
Alexei V. Ivlev
Abstract:
Recent observations of gamma-rays with the Fermi Large Area Telescope (LAT) in the direction of the inner Galaxy revealed a mysterious GeV excess. Its intensity is significantly above predictions of the standard model of cosmic rays (CRs) generation and propagation with a peak in the spectrum around a few GeV. Popular interpretations of this excess are due to either spherically distributed annihil…
▽ More
Recent observations of gamma-rays with the Fermi Large Area Telescope (LAT) in the direction of the inner Galaxy revealed a mysterious GeV excess. Its intensity is significantly above predictions of the standard model of cosmic rays (CRs) generation and propagation with a peak in the spectrum around a few GeV. Popular interpretations of this excess are due to either spherically distributed annihilating dark matter (DM) or abnormal population of millisecond pulsars. We suggested an alternative explanation of the excess through the CR interactions with molecular clouds in the Galactic Center (GC) region. We assumed that the excess could be imitated by the emission of molecular clouds with depleted density of CRs with energies below ~ 10 GeV inside. A novelty of our work is in detailed elaboration of the depletion mechanism of CRs with the mentioned energies through the "barrier" near the cloud edge formed by the self-excited MHD turbulence. Such depletion of CRs inside the clouds may be a reason of deficit of gamma rays from the Central Molecular Zone (CMZ) at energies below few GeV. This in turn changes the ratio between various emission components at those energies, and may potentially absorb the GeV excess by simple renormalization of key components.
△ Less
Submitted 9 August, 2021;
originally announced August 2021.
-
Capabilities of the GAMMA-400 gamma-ray telescope to detect gamma-ray bursts from lateral directions
Authors:
A A Leonov,
A M Galper,
N P Topchiev,
I V Arkhangelskaja,
A I Arkhangelskiy,
A V Bakaldin,
I V Chernysheva,
O D Dalkarov,
A E Egorov,
M D Kheymits,
M G Korotkov,
A G Malinin,
A G Mayorov,
V V Mikhailov,
A V Mikhailova,
P Yu Minaev,
N Yu Pappe,
P Picozza,
R Sparvoli,
Yu I Stozhkov,
S I Suchkov,
Yu T Yurkin
Abstract:
The currently develo** space-based gamma-ray telescope GAMMA-400 will measure the gamma-ray and electron + positron fluxes using the main top-down aperture in the energy range from ~20 MeV to several TeV in a highly elliptic orbit (without shading the telescope by the Earth and outside the radiation belts) continuously for a long time. The instrument will provide fundamentally new data on discre…
▽ More
The currently develo** space-based gamma-ray telescope GAMMA-400 will measure the gamma-ray and electron + positron fluxes using the main top-down aperture in the energy range from ~20 MeV to several TeV in a highly elliptic orbit (without shading the telescope by the Earth and outside the radiation belts) continuously for a long time. The instrument will provide fundamentally new data on discrete gamma-ray sources, gamma-ray bursts (GRBs), sources and propagation of Galactic cosmic rays and signatures of dark matter due to its unique angular and energy resolutions in the wide energy range. The gamma-ray telescope consists of the anticoincidence system (AC), the converter-tracker (C), the time-of-flight system (S1 and S2), the position-sensitive and electromagnetic calorimeters (CC1 and CC2), scintillation detectors (S3 and S4) located above and behind the CC2 calorimeter and lateral detectors (LD) located around the CC2 calorimeter. In this paper, the capabilities of the GAMMA-400 gamma-ray telescope to measure fluxes of GRBs from lateral directions of CC2 are analyzed using Monte-Carlo simulations. The analysis is based on off-line second-level trigger construction using signals from S3, CC2, S4 and LD detectors. For checking the numerical algorithm the data from space-based GBM and LAT instruments of the Fermi experiment are used, namely, three long bursts: GRB 080916C, GRB 090902B, GRB 090926A and one short burst GRB 090510A. The obtained results allow us to conclude that from lateral directions the GAMMA-400 space-based gamma-ray telescope will reliably measure the spectra of bright GRBs in the energy range from ~10 to ~100 MeV with the on-axis effective area of about 0.13 m2 for each of the four sides of CC2 and total field of view of about 6 sr.
△ Less
Submitted 21 January, 2022; v1 submitted 12 March, 2021;
originally announced March 2021.
-
Involutive MCMC: a Unifying Framework
Authors:
Kirill Neklyudov,
Max Welling,
Evgenii Egorov,
Dmitry Vetrov
Abstract:
Markov Chain Monte Carlo (MCMC) is a computational approach to fundamental problems such as inference, integration, optimization, and simulation. The field has developed a broad spectrum of algorithms, varying in the way they are motivated, the way they are applied and how efficiently they sample. Despite all the differences, many of them share the same core principle, which we unify as the Involu…
▽ More
Markov Chain Monte Carlo (MCMC) is a computational approach to fundamental problems such as inference, integration, optimization, and simulation. The field has developed a broad spectrum of algorithms, varying in the way they are motivated, the way they are applied and how efficiently they sample. Despite all the differences, many of them share the same core principle, which we unify as the Involutive MCMC (iMCMC) framework. Building upon this, we describe a wide range of MCMC algorithms in terms of iMCMC, and formulate a number of "tricks" which one can use as design principles for develo** new MCMC algorithms. Thus, iMCMC provides a unified view of many known MCMC algorithms, which facilitates the derivation of powerful extensions. We demonstrate the latter with two examples where we transform known reversible MCMC algorithms into more efficient irreversible ones.
△ Less
Submitted 30 June, 2020;
originally announced June 2020.
-
Bayesian Generative Models for Knowledge Transfer in MRI Semantic Segmentation Problems
Authors:
Anna Kuzina,
Evgenii Egorov,
Evgeny Burnaev
Abstract:
Automatic segmentation methods based on deep learning have recently demonstrated state-of-the-art performance, outperforming the ordinary methods. Nevertheless, these methods are inapplicable for small datasets, which are very common in medical problems. To this end, we propose a knowledge transfer method between diseases via the Generative Bayesian Prior network. Our approach is compared to a pre…
▽ More
Automatic segmentation methods based on deep learning have recently demonstrated state-of-the-art performance, outperforming the ordinary methods. Nevertheless, these methods are inapplicable for small datasets, which are very common in medical problems. To this end, we propose a knowledge transfer method between diseases via the Generative Bayesian Prior network. Our approach is compared to a pre-train approach and random initialization and obtains the best results in terms of Dice Similarity Coefficient metric for the small subsets of the Brain Tumor Segmentation 2018 database (BRATS2018).
△ Less
Submitted 27 May, 2020; v1 submitted 26 May, 2020;
originally announced May 2020.
-
Dark matter searches by the planned gamma-ray telescope GAMMA-400
Authors:
Andrey E. Egorov,
Nikolay P. Topchiev,
Arkadiy M. Galper,
Oleg D. Dalkarov,
Alexey A. Leonov,
Sergey I. Suchkov,
Yuriy T. Yurkin
Abstract:
Our paper reviews the planned space-based gamma-ray telescope GAMMA-400 and evaluates in details its opportunities in the field of dark matter (DM) indirect searches. We estimated GAMMA-400 mean sensitivity to the diphoton DM annihilation cross section in the Galactic center for DM particle masses in the range of 1-500 GeV. We obtained the sensitivity gain at least by 1.2-1.5 times (depending on D…
▽ More
Our paper reviews the planned space-based gamma-ray telescope GAMMA-400 and evaluates in details its opportunities in the field of dark matter (DM) indirect searches. We estimated GAMMA-400 mean sensitivity to the diphoton DM annihilation cross section in the Galactic center for DM particle masses in the range of 1-500 GeV. We obtained the sensitivity gain at least by 1.2-1.5 times (depending on DM particle mass) with respect to the expected constraints from 12 years of observations by Fermi-LAT for the case of Einasto DM density profile. The joint analysis of the data from both telescopes may yield the gain up to 1.8-2.3 times. Thus the sensitivity reaches the level of annihilation cross section $\langle σv \rangle_{γγ}(m_χ=100~\mbox{GeV})\approx 10^{-28}$ cm$^3$/s. This will allow us to test the hypothesized narrow lines predicted by specific DM models, particularly the recently proposed pseudo-Goldstone boson DM model. We also considered the decaying DM - in this case the joint analysis may yield the sensitivity gain up to 1.1-2.0 times reaching the level of DM lifetime $τ_{γν}(m_χ=100~\mbox{GeV}) \approx 2\cdot 10^{29}$ s. We estimated the GAMMA-400 sensitivity to axion-like particle (ALP) parameters by a potential observation of the supernova explosion in the Local Group. This is very sensitive probe of ALPs reaching the level of ALP-photon coupling constant $g_{aγ} \sim 10^{-13}~\mbox{GeV}^{-1}$ for ALP masses $m_a \lesssim 1$ neV. We also calculated the sensitivity to ALPs by constraining the modulations in the spectra of the Galactic gamma-ray pulsars due to possible ALP-photon conversion. GAMMA-400 is expected to be more sensitive than the CAST helioscope for ALP masses $m_a \approx (1-10)$ neV reaching $g_{aγ}^{min} \approx 2\cdot 10^{-11}~\mbox{GeV}^{-1}$. Other potentially interesting targets and candidates are briefly considered too.
△ Less
Submitted 25 November, 2020; v1 submitted 18 May, 2020;
originally announced May 2020.
-
BooVAE: Boosting Approach for Continual Learning of VAE
Authors:
Anna Kuzina,
Evgenii Egorov,
Evgeny Burnaev
Abstract:
Variational autoencoder (VAE) is a deep generative model for unsupervised learning, allowing to encode observations into the meaningful latent space. VAE is prone to catastrophic forgetting when tasks arrive sequentially, and only the data for the current one is available. We address this problem of continual learning for VAEs. It is known that the choice of the prior distribution over the latent…
▽ More
Variational autoencoder (VAE) is a deep generative model for unsupervised learning, allowing to encode observations into the meaningful latent space. VAE is prone to catastrophic forgetting when tasks arrive sequentially, and only the data for the current one is available. We address this problem of continual learning for VAEs. It is known that the choice of the prior distribution over the latent space is crucial for VAE in the non-continual setting. We argue that it can also be helpful to avoid catastrophic forgetting. We learn the approximation of the aggregated posterior as a prior for each task. This approximation is parametrised as an additive mixture of distributions induced by encoder evaluated at trainable pseudo-inputs. We use a greedy boosting-like approach with entropy regularisation to learn the components. This method encourages components diversity, which is essential as we aim at memorising the current task with the fewest components possible. Based on the learnable prior, we introduce an end-to-end approach for continual learning of VAEs and provide empirical studies on commonly used benchmarks (MNIST, Fashion MNIST, NotMNIST) and CelebA datasets. For each dataset, the proposed method avoids catastrophic forgetting in a fully automatic way.
△ Less
Submitted 3 November, 2021; v1 submitted 30 August, 2019;
originally announced August 2019.
-
Bayesian Generative Models for Knowledge Transfer in MRI Semantic Segmentation Problems
Authors:
Anna Kuzina,
Evgenii Egorov,
Evgeny Burnaev
Abstract:
Automatic segmentation methods based on deep learning have recently demonstrated state-of-the-art performance, outperforming the ordinary methods. Nevertheless, these methods are inapplicable for small datasets, which are very common in medical problems. To this end, we propose a knowledge transfer method between diseases via the Generative Bayesian Prior network. Our approach is compared to a pre…
▽ More
Automatic segmentation methods based on deep learning have recently demonstrated state-of-the-art performance, outperforming the ordinary methods. Nevertheless, these methods are inapplicable for small datasets, which are very common in medical problems. To this end, we propose a knowledge transfer method between diseases via the Generative Bayesian Prior network. Our approach is compared to a pre-train approach and random initialization and obtains the best results in terms of Dice Similarity Coefficient metric for the small subsets of the Brain Tumor Segmentation 2018 database (BRATS2018).
△ Less
Submitted 15 August, 2019;
originally announced August 2019.
-
The Implicit Metropolis-Hastings Algorithm
Authors:
Kirill Neklyudov,
Evgenii Egorov,
Dmitry Vetrov
Abstract:
Recent works propose using the discriminator of a GAN to filter out unrealistic samples of the generator. We generalize these ideas by introducing the implicit Metropolis-Hastings algorithm. For any implicit probabilistic model and a target distribution represented by a set of samples, implicit Metropolis-Hastings operates by learning a discriminator to estimate the density-ratio and then generati…
▽ More
Recent works propose using the discriminator of a GAN to filter out unrealistic samples of the generator. We generalize these ideas by introducing the implicit Metropolis-Hastings algorithm. For any implicit probabilistic model and a target distribution represented by a set of samples, implicit Metropolis-Hastings operates by learning a discriminator to estimate the density-ratio and then generating a chain of samples. Since the approximation of density ratio introduces an error on every step of the chain, it is crucial to analyze the stationary distribution of such chain. For that purpose, we present a theoretical result stating that the discriminator loss upper bounds the total variation distance between the target distribution and the stationary distribution. Finally, we validate the proposed algorithm both for independent and Markov proposals on CIFAR-10 and CelebA datasets.
△ Less
Submitted 9 June, 2019;
originally announced June 2019.
-
Usage of multiple RTL features for Earthquake prediction
Authors:
P. Proskura,
A. Zaytsev,
I. Braslavsky,
E. Egorov,
E. Burnaev
Abstract:
We construct a classification model that predicts if an earthquake with the magnitude above a threshold will take place at a given location in a time range 30-180 days from a given moment of time. A common approach is to use expert forecasts based on features like Region-Time-Length (RTL) characteristics. The proposed approach uses machine learning on top of multiple RTL features to take into acco…
▽ More
We construct a classification model that predicts if an earthquake with the magnitude above a threshold will take place at a given location in a time range 30-180 days from a given moment of time. A common approach is to use expert forecasts based on features like Region-Time-Length (RTL) characteristics. The proposed approach uses machine learning on top of multiple RTL features to take into account effects at various scales and to improve prediction accuracy. For historical data about Japan earthquakes 1992-2005 and predictions at locations given in this database the best model has precision up to ~ 0.95 and recall up to ~ 0.98.
△ Less
Submitted 26 May, 2019;
originally announced May 2019.
-
MaxEntropy Pursuit Variational Inference
Authors:
Evgenii Egorov,
Kirill Neklydov,
Ruslan Kostoev,
Evgeny Burnaev
Abstract:
One of the core problems in variational inference is a choice of approximate posterior distribution. It is crucial to trade-off between efficient inference with simple families as mean-field models and accuracy of inference. We propose a variant of a greedy approximation of the posterior distribution with tractable base learners. Using Max-Entropy approach, we obtain a well-defined optimization pr…
▽ More
One of the core problems in variational inference is a choice of approximate posterior distribution. It is crucial to trade-off between efficient inference with simple families as mean-field models and accuracy of inference. We propose a variant of a greedy approximation of the posterior distribution with tractable base learners. Using Max-Entropy approach, we obtain a well-defined optimization problem. We demonstrate the ability of the method to capture complex multimodal posterior via continual learning setting for neural networks.
△ Less
Submitted 19 May, 2019;
originally announced May 2019.
-
Metropolis-Hastings view on variational inference and adversarial training
Authors:
Kirill Neklyudov,
Evgenii Egorov,
Pavel Shvechikov,
Dmitry Vetrov
Abstract:
A significant part of MCMC methods can be considered as the Metropolis-Hastings (MH) algorithm with different proposal distributions. From this point of view, the problem of constructing a sampler can be reduced to the question - how to choose a proposal for the MH algorithm? To address this question, we propose to learn an independent sampler that maximizes the acceptance rate of the MH algorithm…
▽ More
A significant part of MCMC methods can be considered as the Metropolis-Hastings (MH) algorithm with different proposal distributions. From this point of view, the problem of constructing a sampler can be reduced to the question - how to choose a proposal for the MH algorithm? To address this question, we propose to learn an independent sampler that maximizes the acceptance rate of the MH algorithm, which, as we demonstrate, is highly related to the conventional variational inference. For Bayesian inference, the proposed method compares favorably against alternatives to sample from the posterior distribution. Under the same approach, we step beyond the scope of classical MCMC methods and deduce the Generative Adversarial Networks (GANs) framework from scratch, treating the generator as the proposal and the discriminator as the acceptance test. On real-world datasets, we improve Frechet Inception Distance and Inception Score, using different GANs as a proposal distribution for the MH algorithm. In particular, we demonstrate improvements of recently proposed BigGAN model on ImageNet.
△ Less
Submitted 9 June, 2019; v1 submitted 16 October, 2018;
originally announced October 2018.
-
Precise tracking of vaccine-responding T-cell clones reveals convergent and personalized response in identical twins
Authors:
Mikhail V. Pogorelyy,
Anastasia A. Minervina,
Maximilian Puelma Touzel,
Anastasiia L. Sycheva,
Ekaterina A. Komech,
Elena I. Kovalenko,
Galina G. Karganova,
Evgeniy S. Egorov,
Alexander Yu. Komkov,
Dmitriy M. Chudakov,
Ilgar Z. Mamedov,
Thierry Mora,
Aleksandra M. Walczak,
Yuri B. Lebedev
Abstract:
T-cell receptor (TCR) repertoire data contain information about infections that could be used in disease diagnostics and vaccine development, but extracting that information remains a major challenge. Here we developed a statistical framework to detect TCR clone proliferation and contraction from longitudinal repertoire data. We applied this framework to data from three pairs of identical twins im…
▽ More
T-cell receptor (TCR) repertoire data contain information about infections that could be used in disease diagnostics and vaccine development, but extracting that information remains a major challenge. Here we developed a statistical framework to detect TCR clone proliferation and contraction from longitudinal repertoire data. We applied this framework to data from three pairs of identical twins immunized with the yellow fever vaccine. We identified 500-1500 responding TCRs in each donor and validated them using three independent assays. While the responding TCRs were mostly private, albeit with higher overlap between twins, they could be well predicted using a classifier based on sequence similarity. Our method can also be applied to samples obtained post-infection, making it suitable for systematic discovery of new infection-specific TCRs in the clinic.
△ Less
Submitted 12 April, 2018;
originally announced April 2018.
-
Detectability of dark matter subhalos by means of the GAMMA-400 telescope
Authors:
A. E. Egorov,
A. M. Galper,
N. P. Topchiev,
A. A. Leonov,
S. I. Suchkov,
M. D. Kheymits,
Yu. T. Yurkin
Abstract:
We investigated the detectability of Galactic subhalos with masses $(10^6-10^9)M_{\odot}$ formed by annihilating WIMP dark matter by the planned GAMMA-400 gamma-ray telescope. The inner structure of dark matter subhalos and their distribution in the Galaxy were taken from corresponding simulations. We showed that the expected gamma-ray flux from subhalos strongly depends on WIMP mass and subhalo c…
▽ More
We investigated the detectability of Galactic subhalos with masses $(10^6-10^9)M_{\odot}$ formed by annihilating WIMP dark matter by the planned GAMMA-400 gamma-ray telescope. The inner structure of dark matter subhalos and their distribution in the Galaxy were taken from corresponding simulations. We showed that the expected gamma-ray flux from subhalos strongly depends on WIMP mass and subhalo concentration, but less strongly depends on the subhalo mass. In an optimistic case we may expect the flux of 10-100 ph/year above 100 MeV from the closest and most massive subhalos, which would be detectable sources for GAMMA-400. However, resolving the inner structure of subhalos might be possible only by the joint analysis of the future GAMMA-400 data and data from other telescopes due to smallness of fluxes. Also we considered the recent subhalo candidates 3FGL J2212.5+0703 and J1924.8-1034 within the framework of our model. We concluded that it is very unlikely that these sources belong to the subhalo population.
△ Less
Submitted 26 April, 2018; v1 submitted 6 October, 2017;
originally announced October 2017.
-
High-energy gamma-ray studying with GAMMA-400
Authors:
N. P. Topchiev,
A. M. Galper,
V. Bonvicini,
I. V. Arkhangelskaja,
A. I. Arkhangelskiy,
A. V. Bakaldin,
S. G. Bobkov,
O. D. Dalkarov,
A. E. Egorov,
Yu. V. Gusakov,
B. I. Hnatyk,
V. V. Kadilin,
M. D. Kheymits,
V. E. Korepanov,
A. A. Leonov,
V. V. Mikhailov,
A. A. Moiseev,
I. V. Moskalenko,
P. Yu. Naumov,
P. Picozza,
M. F. Runtso,
O. V. Serdin,
R. Sparvoli,
P. Spillantini,
Yu. I. Stozhkov
, et al. (4 additional authors not shown)
Abstract:
Extraterrestrial gamma-ray astronomy is now a source of new knowledge in the fields of astrophysics, cosmic-ray physics, and the nature of dark matter. The next absolutely necessary step in the development of extraterrestrial high-energy gamma-ray astronomy is the improvement of the physical and technical characteristics of gamma-ray telescopes, especially the angular and energy resolutions. Such…
▽ More
Extraterrestrial gamma-ray astronomy is now a source of new knowledge in the fields of astrophysics, cosmic-ray physics, and the nature of dark matter. The next absolutely necessary step in the development of extraterrestrial high-energy gamma-ray astronomy is the improvement of the physical and technical characteristics of gamma-ray telescopes, especially the angular and energy resolutions. Such a new generation telescope will be GAMMA-400. GAMMA-400, currently develo** gamma-ray telescope, together with X-ray telescope will precisely and detailed observe in the energy range of ~20 MeV to ~1000 GeV and 3-30 keV the Galactic plane, especially, Galactic Center, Fermi Bubbles, Crab, Cygnus, etc. The GAMMA- 400 will operate in the highly elliptic orbit continuously for a long time with the unprecedented angular (~0.01° at Eγ = 100 GeV) and energy (~1% at Eγ = 100 GeV) resolutions better than the Fermi-LAT, as well as ground gamma-ray telescopes, by a factor of 5-10. GAMMA-400 will permit to resolve gamma rays from annihilation or decay of dark matter particles, identify many discrete sources (many of which are variable), to clarify the structure of extended sources, to specify the data on the diffuse emission.
△ Less
Submitted 16 July, 2017;
originally announced July 2017.
-
Improvement of the GAMMA-400 physical scheme for precision gamma-ray emission investigations
Authors:
A. A. Leonov,
A. M. Galper,
N. P. Topchiev,
V. Bonvicini,
O. Adriani,
I. V. Arkhangelskaja,
A. I. Arkhangelskiy,
A. V. Bakaldin,
S. G. Bobkov,
M. Boezio,
O. D. Dalkarov,
A. E. Egorov,
N. A. Glushkov,
M. S. Gorbunov,
Yu. V. Gusakov,
B. I. Hnatyk,
V. V. Kadilin,
V. A. Kaplin,
M. D. Kheymits,
V. E. Korepanov,
F. Longo,
V. V. Mikhailov,
E. Mocchiutti,
A. A. Moiseev,
I. V. Moskalenko
, et al. (12 additional authors not shown)
Abstract:
The main goal for the GAMMA-400 gamma-ray telescope mission is to perform a sensitive search for signatures of dark matter particles in high-energy gamma-ray emission. Measurements will also concern the following scientific goals: detailed study of the Galactic center region, investigation of point and extended gamma-ray sources, studies of the energy spectra of Galactic and extragalactic diffuse…
▽ More
The main goal for the GAMMA-400 gamma-ray telescope mission is to perform a sensitive search for signatures of dark matter particles in high-energy gamma-ray emission. Measurements will also concern the following scientific goals: detailed study of the Galactic center region, investigation of point and extended gamma-ray sources, studies of the energy spectra of Galactic and extragalactic diffuse emissions. To perform these measurements the GAMMA-400 gamma-ray telescope possesses unique physical characteristics for energy range from ~20 MeV to ~1000 GeV in comparison with previous and current space and ground-based experiments. The major advantage of the GAMMA-400 instrument is excellent angular and energy resolutions for gamma-rays above 10 GeV. The gamma-ray telescope angular and energy resolutions for the main aperture at 100-GeV gamma rays are ~0.01 deg and ~1%, respectively. The special goal is to improve physical characteristics in the low- energy range from ~20 MeV to 100 MeV. Minimizing the amount of dead matter in the telescope aperture allows us to obtain the angular and energy resolutions better in this range than in current space missions. The gamma-ray telescope angular resolution at 50-MeV gamma rays is better than 5 deg and energy resolution is ~10%. We report the method providing these results.
△ Less
Submitted 28 December, 2016;
originally announced December 2016.
-
Dark matter implications of the WMAP-Planck Haze
Authors:
Andrey E. Egorov,
Jennifer M. Gaskins,
Elena Pierpaoli,
Davide Pietrobon
Abstract:
Gamma rays and microwave observations of the Galactic Center and surrounding areas indicate the presence of anomalous emission, whose origin remains ambiguous. The possibility of dark matter (DM) annihilation explaining both signals through prompt emission at gamma-rays and secondary emission at microwave frequencies from interactions of high-energy electrons produced in annihilation with the Gala…
▽ More
Gamma rays and microwave observations of the Galactic Center and surrounding areas indicate the presence of anomalous emission, whose origin remains ambiguous. The possibility of dark matter (DM) annihilation explaining both signals through prompt emission at gamma-rays and secondary emission at microwave frequencies from interactions of high-energy electrons produced in annihilation with the Galactic magnetic fields has attracted much interest in recent years. We investigate the DM interpretation of the Galactic Center gamma-ray excess by searching for the associated synchrotron in the WMAP-Planck data. Considering various magnetic field and cosmic-ray propagation models, we predict the synchrotron emission due to DM annihilation in our Galaxy, and compare it with the WMAP-Planck data at 23-70GHz. In addition to standard microwave foregrounds, we separately model the microwave counterpart to the Fermi Bubbles and the signal due to DM, and use component separation techniques to extract the signal associated with each template from the total emission. We confirm the presence of the Haze at the level of 7% of the total sky intensity at 23GHz in our chosen region of interest, with a harder spectrum $I \sim ν^{-0.8}$ than the synchrotron from regular cosmic-ray electrons. The data do not show a strong preference towards fitting the Haze by either the Bubbles or DM emission only. Inclusion of both components provides a better fit with a DM contribution to the Haze emission of 20% at 23GHz, however, due to significant uncertainties in foreground modeling, we do not consider this a clear detection of a DM signal. We set robust upper limits on the annihilation cross section by ignoring foregrounds, and also report best-fit DM annihilation parameters obtained from a complete template analysis. We conclude that the WMAP-Planck data are consistent with a DM interpretation of the gamma-ray excess.
△ Less
Submitted 26 April, 2016; v1 submitted 17 September, 2015;
originally announced September 2015.
-
Constraints on dark matter annihilation by radio observations of M31
Authors:
A. E. Egorov,
E. Pierpaoli
Abstract:
We used radio observations of the neighbour galaxy M31 in order to put constraints on dark matter particle mass and annihilation cross section. Dark matter annihilation in M31 halo produces highly energetic leptons, which emit synchrotron radiation on radio frequencies in the galactic magnetic field. We predicted expected radio fluxes for the two annihilation channels: χχ-> bb* and χχ-> τ^+τ^-. We…
▽ More
We used radio observations of the neighbour galaxy M31 in order to put constraints on dark matter particle mass and annihilation cross section. Dark matter annihilation in M31 halo produces highly energetic leptons, which emit synchrotron radiation on radio frequencies in the galactic magnetic field. We predicted expected radio fluxes for the two annihilation channels: χχ-> bb* and χχ-> τ^+τ^-. We then compared them with available data on the central radio emission of M31 as observed by four radio surveys: VLSS (74 MHz), WENSS (325 MHz), NVSS (1400 MHz) and GB6 (4850 MHz). Assuming a standard NFW dark matter density profile and a conservative magnetic field distribution inside the Andromeda galaxy, we find that the thermal relic annihilation cross section <σv> = 3*10^{-26} cm^3/s or higher are only allowed for WIMP masses greater than 100 GeV and 55 GeV for annihilation into bb* and τ^+τ^- respectively. Taking into account potential uncertainties in the distributions of DM density and magnetic field, the mentioned WIMP limiting masses can be as low as 23 GeV for both channels, and as high as 280 and 130 GeV for annihilation into bb* and τ^+τ^- respectively. These mass values exceed the best up-to-day known constraints from Fermi gamma observations: 40 GeV and 19 GeV respectively [A.Geringer-Sameth and S.M.Koushiappas, Phys. Rev. Lett. 107, 241303 (2011)]. Precise measurements of the magnetic field in the relevant region and better reconstruction of the DM density profile of M31 will be able to reduce the uncertainties of our exclusion limits.
△ Less
Submitted 1 April, 2013;
originally announced April 2013.
-
On the possible observational manifestation of supernova shock impact on the neutron star magnetosphere
Authors:
A. E. Egorov,
K. A. Postnov
Abstract:
Impact of supernova explosion on the neutron star magnetosphere in a massive binary system is considered. The supernova shock striking the NS magnetosphere filled with plasma can lead to the formation of a magnetospheric tail with significant magnetic energy. The magnetic field reconnection in the current sheet formed can convert the magnetic energy stored in the tail into kinetic energy of acce…
▽ More
Impact of supernova explosion on the neutron star magnetosphere in a massive binary system is considered. The supernova shock striking the NS magnetosphere filled with plasma can lead to the formation of a magnetospheric tail with significant magnetic energy. The magnetic field reconnection in the current sheet formed can convert the magnetic energy stored in the tail into kinetic energy of accelerated charged particles. Plasma instabilities excited by beams of relativistic particles can lead to the formation of a short pulse of coherent radio emission with parameters similar to those of the observed bright extragalactic millisecond radio burst (Lorimer et al. 2007).
△ Less
Submitted 13 October, 2008;
originally announced October 2008.
-
Analysis of the Formation of Structures and Chaotic Dynamics in a Nonrelativistic Electron Beam with a Virtual Cathode in the Presence of a Decelerating Field
Authors:
E. N. Egorov,
Yu. A. Kalinin,
A. A. Koronovskii,
Yu. I. Levin,
A. E. Hramov
Abstract:
A nonrelativistic electron beam with a virtual cathode situated in the diode gap with a decelerating field is experimentally and theoretically studied. A 1D model of the electron beam in the presence of a decelerating field is constructed. Nonlinear nonstationary processes in this system are investigated by means of numerical analysis of the model. The processes are described and interpreted wit…
▽ More
A nonrelativistic electron beam with a virtual cathode situated in the diode gap with a decelerating field is experimentally and theoretically studied. A 1D model of the electron beam in the presence of a decelerating field is constructed. Nonlinear nonstationary processes in this system are investigated by means of numerical analysis of the model. The processes are described and interpreted with regard to formation and interaction of structures in the electron flow. The theoretical results are qualitatively confirmed by the experimental data showing that the system under study can be considered as a promising controlled source of chaotic oscillations in the microwave range.
△ Less
Submitted 27 January, 2007;
originally announced January 2007.
-
Investigation of the Chaotic Dynamics of an Electron Beam with a Virtual Cathode in an External Magnetic Field
Authors:
E. N. Egorov,
A. E. Hramov
Abstract:
The effect of the strength of the focusing magnetic field on chaotic dynamic processes occurring inan electron beam with a virtual cathode, as well as on the processes whereby the structures form in the beamand interact with each other, is studied by means of two-dimensional numerical simulations based on solving a self-consistent set of Vlasov-Maxwell equations. It is shown that, as the focusin…
▽ More
The effect of the strength of the focusing magnetic field on chaotic dynamic processes occurring inan electron beam with a virtual cathode, as well as on the processes whereby the structures form in the beamand interact with each other, is studied by means of two-dimensional numerical simulations based on solving a self-consistent set of Vlasov-Maxwell equations. It is shown that, as the focusing magnetic field is decreased,the dynamics of an electron beam with a virtual cathode becomes more complicated due to the formation andinteraction of spatio-temporal longitudinal and transverse structures in the interaction region of a vircator. The optimum efficiency of the interaction of an electron beam with the electromagnetic field of the vircator isachieved at a comparatively weak external magnetic field and is determined by the fundamentally two-dimensional nature of the motion of the beam electrons near the virtual cathode.
△ Less
Submitted 27 June, 2006;
originally announced June 2006.
-
Microwave Generation Power in a Nonrelativistic Electron Beam with Virtual Cathode in a Retarding Electric Field
Authors:
E. N. Egorov,
Yu. A. Kalinin,
A. A. Koronovskii,
A. E. Hramov,
M. Yu. Morozov
Abstract:
The power of microwave generation in a nonrelativistic electron beam with virtual cathode formed in a static retarding electric field (low-voltage vircator system) has been studied experimentally and by means of numerical simulation within the framework of a one-dimensional theory. The limits of applicability of the one-dimensional theory have been experimentally determined.
The power of microwave generation in a nonrelativistic electron beam with virtual cathode formed in a static retarding electric field (low-voltage vircator system) has been studied experimentally and by means of numerical simulation within the framework of a one-dimensional theory. The limits of applicability of the one-dimensional theory have been experimentally determined.
△ Less
Submitted 19 May, 2006;
originally announced May 2006.
-
Analysis of Transient Processes in a Radiophysical Flow System
Authors:
E. N. Egorov,
A. A. Koronovskii,
A. E. Hramov
Abstract:
Transient processes in a third-order radiophysical flow system are studied and a map of the transient process duration versus initial conditions is constructed and analyzed. The results are compared to the arrangement of submanifolds of the stable and unstable cycles in the Poincare section of the system studied.
Transient processes in a third-order radiophysical flow system are studied and a map of the transient process duration versus initial conditions is constructed and analyzed. The results are compared to the arrangement of submanifolds of the stable and unstable cycles in the Poincare section of the system studied.
△ Less
Submitted 31 January, 2006;
originally announced January 2006.
-
Experimental and Theoretical Investigations of Stochastic Oscillatory Phenomena in a Nonrelativistic Electron Beam with a Virtual Cathode
Authors:
Yu. A. Kalinin,
A. A. Koronovskii,
A. E. Hramov,
E. N. Egorov,
R. A. Filatov
Abstract:
Results are presented from experimental investigations of oscillatory phenomena in an electron beam with a virtual cathode in a diode gap with a decelerating field. Experiments have revealed a stochastic broadband generation of the microwave oscillations of a virtual cathode in a decelerating field. Numerical simulations based on a simple one-dimensional model have shown that the onset of the st…
▽ More
Results are presented from experimental investigations of oscillatory phenomena in an electron beam with a virtual cathode in a diode gap with a decelerating field. Experiments have revealed a stochastic broadband generation of the microwave oscillations of a virtual cathode in a decelerating field. Numerical simulations based on a simple one-dimensional model have shown that the onset of the stochastic generation and the broadening of the oscillation spectrum with increasing beam deceleration rate are governed by the processes of regrou** of the electrons in a beam with a virtual cathode.
△ Less
Submitted 28 January, 2006;
originally announced January 2006.
-
Discovery of a Planar Waveguide for an X-Ray Radiation
Authors:
V. Egorov,
E. Egorov
Abstract:
A simple model of X-Ray standing waves (XSW) formation in the slit of a planar waveguide of X-Ray radiation beam for the angle area restricted by the critical total reflection angle is developed. It is shown that the model is true for a case of the Bragg reflection. The conditions required for XSW to appear in the space between two polished parallel plane plates are formulated and a slit size in…
▽ More
A simple model of X-Ray standing waves (XSW) formation in the slit of a planar waveguide of X-Ray radiation beam for the angle area restricted by the critical total reflection angle is developed. It is shown that the model is true for a case of the Bragg reflection. The conditions required for XSW to appear in the space between two polished parallel plane plates are formulated and a slit size interval conforming to these conditions is evaluated. A mechanism of a XSW intensity decrease in a planar waveguide is proposed. This mechanism explains a high efficiency of slitless collimator application for the transportation of narrow X-Ray beams.
Some recommendations on the application of the planar X-Ray waveguide in X-Ray structural and spectral studies of surface are presented.
△ Less
Submitted 14 January, 2001;
originally announced January 2001.
