-
Looking around you: external information enhances representations for event sequences
Authors:
Maria Kovaleva,
Petr Sokerin,
Sofia Krehova,
Alexey Zaytsev
Abstract:
Representation learning produces models in different domains, such as store purchases, client transactions, and general people's behaviour. However, such models for sequential data usually process a single sequence, ignoring context from other relevant ones, even in domains with rapidly changing external environments like finance or misguiding the prediction for a user with no recent events.
We…
▽ More
Representation learning produces models in different domains, such as store purchases, client transactions, and general people's behaviour. However, such models for sequential data usually process a single sequence, ignoring context from other relevant ones, even in domains with rapidly changing external environments like finance or misguiding the prediction for a user with no recent events.
We are the first to propose a method that aggregates information from multiple user representations augmenting a specific user one for a scenario of multiple co-occurring event sequences. Our study considers diverse aggregation approaches, ranging from simple pooling techniques to trainable attention-based approaches, especially Kernel attention aggregation, that can highlight more complex information flow from other users. The proposed method operates atop an existing encoder and supports its efficient fine-tuning. Across considered datasets of financial transactions and downstream tasks, Kernel attention improves ROC AUC scores, both with and without fine-tuning, while mean pooling yields a smaller but still significant gain.
△ Less
Submitted 14 February, 2025;
originally announced February 2025.
-
Belyi Function Decompositions for The Icosahedron of Genus 4
Authors:
Natalia Amburg,
Mariia Kovaleva
Abstract:
The icosahedron $I_4$ of genus 4 is a dessin d'enfant embedded in Bring's curve $\mathcal{B}$. The dessin $I_4$ is related in some sense to a regular icosahedron $I_0$ embedded in the complex Riemann sphere. In particular, decompositions of Belyi functions $β_{I_0}: \mathbb{CP}^1 \rightarrow \mathbb{CP}^1$ and $β_{I_4}: \mathcal{B} \rightarrow \mathbb{CP}^1$ for $I_0$ and $I_4$ have the same latti…
▽ More
The icosahedron $I_4$ of genus 4 is a dessin d'enfant embedded in Bring's curve $\mathcal{B}$. The dessin $I_4$ is related in some sense to a regular icosahedron $I_0$ embedded in the complex Riemann sphere. In particular, decompositions of Belyi functions $β_{I_0}: \mathbb{CP}^1 \rightarrow \mathbb{CP}^1$ and $β_{I_4}: \mathcal{B} \rightarrow \mathbb{CP}^1$ for $I_0$ and $I_4$ have the same lattice. The diagram of $β_{I_0}$ decompositions is already known. In the present paper we find $β_{I_4}$ decompositions. Note that $β_{I_0}$ decomposes into rational functions on $\mathbb{C}P^1$, while in case of $β_{I_4}$ we deal with maps between different algebraic curves.
△ Less
Submitted 11 May, 2024;
originally announced May 2024.
-
Learning Transactions Representations for Information Management in Banks: Mastering Local, Global, and External Knowledge
Authors:
Alexandra Bazarova,
Maria Kovaleva,
Ilya Kuleshov,
Evgenia Romanenkova,
Alexander Stepikin,
Alexandr Yugay,
Dzhambulat Mollaev,
Ivan Kireev,
Andrey Savchenko,
Alexey Zaytsev
Abstract:
In today's world, banks use artificial intelligence to optimize diverse business processes, aiming to improve customer experience. Most of the customer-related tasks can be categorized into two groups: 1) local ones, which focus on a client's current state, such as transaction forecasting, and 2) global ones, which consider the general customer behaviour, e.g., predicting successful loan repayment…
▽ More
In today's world, banks use artificial intelligence to optimize diverse business processes, aiming to improve customer experience. Most of the customer-related tasks can be categorized into two groups: 1) local ones, which focus on a client's current state, such as transaction forecasting, and 2) global ones, which consider the general customer behaviour, e.g., predicting successful loan repayment. Unfortunately, maintaining separate models for each task is costly. Therefore, to better facilitate information management, we compared eight state-of-the-art unsupervised methods on 11 tasks in search for a one-size-fits-all solution. Contrastive self-supervised learning methods were demonstrated to excel at global problems, while generative techniques were superior at local tasks. We also introduced a novel approach, which enriches the client's representation by incorporating external information gathered from other clients. Our method outperforms classical models, boosting accuracy by up to 20\%.
△ Less
Submitted 3 February, 2025; v1 submitted 2 April, 2024;
originally announced April 2024.
-
Temperature-induced reversal effects of kink dynamics in carbon nanotube on flat substrate
Authors:
Alexander V. Savin,
Margarita Kovaleva
Abstract:
Carbon nanotubes are nano-objects with quite anisotropic properties, for example the mechanical properties in longitudinal and radial directions differ significantly. This feature of the carbon nanotubes yields many interesting phenomena investigated in last decades. One of them is the ability to form both hollow and collapsed states if the radius of the nanotube is large enough. The transitions b…
▽ More
Carbon nanotubes are nano-objects with quite anisotropic properties, for example the mechanical properties in longitudinal and radial directions differ significantly. This feature of the carbon nanotubes yields many interesting phenomena investigated in last decades. One of them is the ability to form both hollow and collapsed states if the radius of the nanotube is large enough. The transitions between the two states have been also reported. In our study we present single-walled carbon nanotube interacting with a plane substrate and characterize the energy of interaction with the substrate using effective Lennard-Jones-type potential. We show energy of the homogeneous open and collapsed states depending on the radius of the carbon nanotube and report on the bi-stability in some range of the nanotube diameters. Using the molecular-dynamical simulations we look at the evolution of the initial half-opened, half-collapsed state and demonstrate that the transition area from one state to another is spatially localized having features of topological soliton (kink or anti-kink). We show that the value and the direction of the kink propagation speed depend significantly on the nanotube diameter as well as on the temperature of the system. We also discuss the mechanism of the process using a simplified model with asymmetric double-well potential and show the entropic nature of the transition.
△ Less
Submitted 24 October, 2023;
originally announced October 2023.
-
Designing an attack-defense game: how to increase robustness of financial transaction models via a competition
Authors:
Alexey Zaytsev,
Maria Kovaleva,
Alex Natekin,
Evgeni Vorsin,
Valerii Smirnov,
Georgii Smirnov,
Oleg Sidorshin,
Alexander Senin,
Alexander Dudin,
Dmitry Berestnev
Abstract:
Banks routinely use neural networks to make decisions. While these models offer higher accuracy, they are susceptible to adversarial attacks, a risk often overlooked in the context of event sequences, particularly sequences of financial transactions, as most works consider computer vision and NLP modalities.
We propose a thorough approach to studying these risks: a novel type of competition that…
▽ More
Banks routinely use neural networks to make decisions. While these models offer higher accuracy, they are susceptible to adversarial attacks, a risk often overlooked in the context of event sequences, particularly sequences of financial transactions, as most works consider computer vision and NLP modalities.
We propose a thorough approach to studying these risks: a novel type of competition that allows a realistic and detailed investigation of problems in financial transaction data. The participants directly oppose each other, proposing attacks and defenses -- so they are examined in close-to-real-life conditions.
The paper outlines our unique competition structure with direct opposition of participants, presents results for several different top submissions, and analyzes the competition results. We also introduce a new open dataset featuring financial transactions with credit default labels, enhancing the scope for practical research and development.
△ Less
Submitted 19 September, 2024; v1 submitted 22 August, 2023;
originally announced August 2023.
-
What is the SKA-Low Sensitivity for Your Favourite Radio Source ?
Authors:
M. Sokolowski,
S. J. Tingay,
D. B. Davidson,
R. B. Wayth,
D. Ung,
J. W. Broderick,
B. Juswardy,
M. Kovaleva,
G. Macario,
G. Pupillo,
A. Sutinjo
Abstract:
The SKA will be the largest radio astronomy observatory ever built, providing unprecedented sensitivity over a very broad frequency (50 MHz to 15.3 GHz). The SKA-Low (50 - 350 MHz), will be built at the MRO in Western Australia. It will consist of 512 stations each composed of 256 dual-polarised antennas, and the sensitivity of an individual station is pivotal to the performance of the entire SKA-…
▽ More
The SKA will be the largest radio astronomy observatory ever built, providing unprecedented sensitivity over a very broad frequency (50 MHz to 15.3 GHz). The SKA-Low (50 - 350 MHz), will be built at the MRO in Western Australia. It will consist of 512 stations each composed of 256 dual-polarised antennas, and the sensitivity of an individual station is pivotal to the performance of the entire SKA-Low telescope. The answer to the question in the title is, it depends. The sensitivity of a low frequency array, such as an SKA-Low station, depends strongly on the pointing direction of the digitally formed station beam and the local sidereal time (LST), and is different for the two orthogonal polarisations of the antennas. The accurate prediction of the SKA-Low sensitivity in an arbitrary direction in the sky is crucial for future observation planning. We present here a sensitivity calculator for the SKA-Low radio telescope, using a database of pre-computed sensitivity values for two realisations of an SKA-Low station architecture. One realisation uses the log-periodic antennas selected for SKA-Low. The second uses a known benchmark, in the form of the bowtie dipoles of the MWA. Data collected by both stations (deployed at the MRO in 2019) were used to measure their sensitivity at selected frequencies and over at least 24 h intervals, and were compared to the predictions described in this paper. The sensitivity values stored in the SQLite database were pre-computed for the X, Y and Stokes I polarisations in 10 MHz frequency steps, 0.5 hour LST intervals, and 5 degree resolution in pointing directions. The database allows users to estimate the sensitivity of SKA-Low for their favourite object using interactive web-based or command line interface, which can also calculate the sensitivity for arbitrary pointing directions, frequencies, and times without interpolations.
△ Less
Submitted 12 April, 2022;
originally announced April 2022.
-
System equivalent flux density of a low-frequency polarimetric phased array interferometer
Authors:
A. T. Sutinjo,
D. C. X. Ung,
M. Sokolowski,
M. Kovaleva,
S. McSweeney
Abstract:
This paper extends the treatment of system equivalent flux density (SEFD) in Sutinjo, A. T. et al. (2021) (Paper I) to interferometric phased array telescopes. The objective is to develop an SEFD formula involving only the most fundamental assumptions and one that is readily applicable to phased array interferometer radio observations. Then, we aimed at comparing the resultant SEFD expression agai…
▽ More
This paper extends the treatment of system equivalent flux density (SEFD) in Sutinjo, A. T. et al. (2021) (Paper I) to interferometric phased array telescopes. The objective is to develop an SEFD formula involving only the most fundamental assumptions and one that is readily applicable to phased array interferometer radio observations. Then, we aimed at comparing the resultant SEFD expression against the often-used root-mean-square (RMS) SEFD approximation, SEFDrmsI = (1/2)(SEFD^2_XX + SEFD^2_YY)^(1/2) to study the inaccuracy of the SEFDrms.
We take into account all mutual coupling and noise coupling within an array environment (intra-array coupling). This intra-array noise coupling is included in the SEFD expression through the realized noise resistance of the array, which accounts for the system noise. No assumption is made regarding the polarization (or lack thereof) of the sky nor the orthogonality of the antenna elements. The fundamental noise assumption is that, in phasor representation, the real and imaginary components of a given noise source are independent and equally distributed (iid) with zero mean. Noise sources that are mutually correlated and non-iid among themselves are allowed, provided the real and imaginary components of each noise source are iid. The system noise is uncorrelated between array entities separated by a baseline distance, which in the case of the Murchison Widefield Array (MWA) is typically tens of wavelengths or greater. By comparing the resulting SEFD formula to the SEFD_I^rms approximation, we proved that SEFD_I^rms always underestimates the SEFD, which leads to an overestimation of array sensitivity.
△ Less
Submitted 31 January, 2022;
originally announced January 2022.
-
System Equivalent Flux Density of a Polarimetric Tripole Radio Interferometer
Authors:
A. T. Sutinjo,
M. Kovaleva,
Y. Xu
Abstract:
System equivalent flux density (SEFD) is an important figure of merit of a radio telescope. This paper aims to derive a general expression for SEFD of a polarimetric tripole interferometer. The derivation makes only two basic and reasonable assumptions. First, the noise under consideration is zero mean and when expressed in complex phasor domain, has independent and identically distributed (iid) r…
▽ More
System equivalent flux density (SEFD) is an important figure of merit of a radio telescope. This paper aims to derive a general expression for SEFD of a polarimetric tripole interferometer. The derivation makes only two basic and reasonable assumptions. First, the noise under consideration is zero mean and when expressed in complex phasor domain, has independent and identically distributed (iid) real and imaginary components. Correlated and non-identically distributed noise sources are allowed as long as the real and imaginary components remain iid. Second, the system noise is uncorrelated between the elements separated by a baseline distance. The SEFD expression is derived from first principles, that is the standard deviation of the noisy flux estimate in a target direction due to system noise. The resulting SEFD expression is expressed as a simple matrix operation that involves a mixture of the system temperatures of each antenna and the Jones matrix elements. It is not limited to tripoles, but rather, fully extensible to multipole antennas; it is not limited to mutually orthogonal antennas. To illustrate the usefulness of the expression and how the formula is applied, we discuss an example calculation based on a tripole interferometer on lunar orbit for ultra-long wavelengths observation. We compared the SEFD results based on a formula assuming short dipoles and the general expression. As expected, the SEFDs converge at the ultra-long wavelengths where the dipoles are well-approximated as short dipoles. The general SEFD expression can be applied to any multipole antenna systems with arbitrary shapes.
△ Less
Submitted 13 December, 2021;
originally announced December 2021.
-
Sensitivity of a Low-Frequency Polarimetric Radio Interferometer
Authors:
A. T. Sutinjo,
M. Sokolowski,
M. Kovaleva,
D. C. X. Ung,
J. W. Broderick,
R. B. Wayth,
D. B. Davidson,
S. J. Tingay
Abstract:
Aims: This paper aims to derive an expression for the sensitivity of a polarimetric radio interferometer that is valid for all-sky observations of arbitrarily polarized sources, with neither a restriction on FoV nor with any a priori assumption regarding the polarization state of the source. We verify the resulting formula with an all-sky observation using the Murchison Widefield Array (MWA) teles…
▽ More
Aims: This paper aims to derive an expression for the sensitivity of a polarimetric radio interferometer that is valid for all-sky observations of arbitrarily polarized sources, with neither a restriction on FoV nor with any a priori assumption regarding the polarization state of the source. We verify the resulting formula with an all-sky observation using the Murchison Widefield Array (MWA) telescope. Methods: The sensitivity expression is developed from first principles by applying the concept of System Equivalent Flux Density (SEFD) to a polarimetric radio interferometer not by computing $A_e/T_{sys}$. The SEFD is calculated from the standard deviation of the noisy flux density estimate for a target source due to system noise. Results: The SEFD for a polarimetric radio interferometer is generally not $1/\sqrt{2}$ of a single-polarized interferometer as is often assumed for narrow FoV. This assumption can lead to significant errors for a dual-polarized dipole based system, which is common in low-frequency radio astronomy: up to $\sim 15\%$ for a zenith angle (ZA) coverage of $45^\circ$, and up to $\sim45\%$ for $60^\circ$ coverage. The worst case errors occur in the diagonal planes of the dipole for very wide FoV. This is demonstrated through theory, simulation and observations. Furthermore, using the resulting formulation, calculation of the off-zenith sensitivity is straightforward and unambiguous. Conclusions: For wide FoV observations pertinent to low-frequency radio interferometer such as the SKA-Low, the narrow FoV and the single-polarized sensitivity expressions are not correct and should be replaced by the formula derived in this paper.
△ Less
Submitted 14 December, 2020;
originally announced December 2020.
-
Coupled Schrödinger equations as a model of interchain excitation transport in the DNA
Authors:
Margarita Kovaleva,
Leonid Manevitch
Abstract:
In our report we consider two weakly coupled Schrödinger equations as a model of the interchain energy transport in the DNA double-helix. We use the reduction of the Yakushevich-type model considering the torsional dynamics of the DNA. In the previous works only small amplitude excitations and stationary dynamics were investigated, while we focus on the non-stationary dynamics of the double-helix.…
▽ More
In our report we consider two weakly coupled Schrödinger equations as a model of the interchain energy transport in the DNA double-helix. We use the reduction of the Yakushevich-type model considering the torsional dynamics of the DNA. In the previous works only small amplitude excitations and stationary dynamics were investigated, while we focus on the non-stationary dynamics of the double-helix. We consider the system as a model of two weakly interacting DNA strands. Supposing that initially only one of the chains is excited in form of breather we demonstrate the existence of invariant which allows to reduce the order of the problem and consider the system of the phase plane. The analysis provided demonstrates analytical tool for prediction of the periodic interchain excitation transitions of its localization on one of the chains. The technique also takes into account the spreading of the excitations with time.
△ Less
Submitted 9 December, 2020;
originally announced December 2020.
-
Non-conventional phase attractors and repellers in weakly coupled autogenerators with hard excitation
Authors:
Margarita Kovaleva,
Leonid Manevitch,
Valery Pilipchuk
Abstract:
In our earlier studies, we found the effect of non-conventional synchronization, which is a specific type of nonlinear stable beating in the system of two weakly coupled autogenerators with hard excitation given by generalized van der Pol-Duffing characteristics. The corresponding synchronized dynamics are due to a new type of attractor in a reduced phase space of the system. In the present work,…
▽ More
In our earlier studies, we found the effect of non-conventional synchronization, which is a specific type of nonlinear stable beating in the system of two weakly coupled autogenerators with hard excitation given by generalized van der Pol-Duffing characteristics. The corresponding synchronized dynamics are due to a new type of attractor in a reduced phase space of the system. In the present work, we show that, as the strength of nonlinear stiffness and dissipation are changing, the phase portrait undergoes a complicated evolution leading to a quite unexpected appearance of difficult to detect repellers separating a stable limit cycle and equilibrium points in the phase plane. In terms of the original coordinates, the limit cycle associates with nonlinear beatings while the stationary points correspond to the stationary synchronous dynamics similar to the so-called nonlinear local modes.
△ Less
Submitted 20 June, 2017;
originally announced June 2017.
-
Regimes of local energy pulsations in non-linear Klein-Gordon trimer: Higher dimensional analogs of limiting phase trajectories
Authors:
Victor Kislovsky,
Margarita Kovaleva,
Yuli Starosvetsky
Abstract:
In the present paper we consider the dynamics of special class of modulated regimes emerging in the Klein-Gordon trimer. In particular, this study unveils the unique states of resonant energy transfer manifested by the regular energy pulsations localized on the first two elements of the chain. This state is regarded as the higher dimensional analog of limiting phase trajectory corresponding to the…
▽ More
In the present paper we consider the dynamics of special class of modulated regimes emerging in the Klein-Gordon trimer. In particular, this study unveils the unique states of resonant energy transfer manifested by the regular energy pulsations localized on the first two elements of the chain. This state is regarded as the higher dimensional analog of limiting phase trajectory corresponding to the regime of maximal energy transfer in the 2DOF, anharmonic, oscillatory models. We show that in contrast to the limiting phase trajectories, this special state of local energy transfer emerging in the Klein-Gordon trimer undergoes additional transition leading to its complete elimination. By employing the special order reduction procedure, we describe analytically the mechanism of formation and destruction of this peculiar, dynamical state. Finally, we demonstrate the strong correlation between the bifurcations of this state to the series of spontaneous transitions from localized to delocalized energy pulsations undergone by the impulsive response of the Klein-Gordon trimer. Results of the analytical model are well in agreement with these of the numerical simulations.
△ Less
Submitted 23 January, 2016; v1 submitted 6 January, 2016;
originally announced January 2016.
