A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2021; you can also visit the original URL.
The file type is application/pdf.
Filters
Adversarial Attacks on Deep Models for Financial Transaction Records
[article]
2021
arXiv
pre-print
Preserved Fulltext
Machine learning models using transaction records as inputs are popular among financial institutions. The most efficient models use deep-learning architectures similar to those in the NLP community, posing a challenge due to their tremendous number of parameters and limited robustness. In particular, deep-learning models are vulnerable to adversarial attacks: a little change in the input harms the model's output. In this work, we examine adversarial attacks on transaction records data and
arXiv:2106.08361v1
fatcat:l4s5xldrafgater7fv73nl5rme
more »
... es from these attacks. The transaction records data have a different structure than the canonical NLP or time series data, as neighbouring records are less connected than words in sentences, and each record consists of both discrete merchant code and continuous transaction amount. We consider a black-box attack scenario, where the attack doesn't know the true decision model, and pay special attention to adding transaction tokens to the end of a sequence. These limitations provide more realistic scenario, previously unexplored in NLP world. The proposed adversarial attacks and the respective defences demonstrate remarkable performance using relevant datasets from the financial industry. Our results show that a couple of generated transactions are sufficient to fool a deep-learning model. Further, we improve model robustness via adversarial training or separate adversarial examples detection. This work shows that embedding protection from adversarial attacks improves model robustness, allowing a wider adoption of deep models for transaction records in banking and finance.
A Differentiable Language Model Adversarial Attack on Text Classifiers
[article]
2021
arXiv
pre-print
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2021; you can also visit the original URL.
The file type is application/pdf.
Robustness of huge Transformer-based models for natural language processing is an important issue due to their capabilities and wide adoption. One way to understand and improve robustness of these models is an exploration of an adversarial attack scenario: check if a small perturbation of an input can fool a model. Due to the discrete nature of textual data, gradient-based adversarial methods, widely used in computer vision, are not applicable per~se. The standard strategy to overcome this
arXiv:2107.11275v1
fatcat:ava6j4azlzagfm2ibbcks32n7q
more »
... is to develop token-level transformations, which do not take the whole sentence into account. In this paper, we propose a new black-box sentence-level attack. Our method fine-tunes a pre-trained language model to generate adversarial examples. A proposed differentiable loss function depends on a substitute classifier score and an approximate edit distance computed via a deep learning model. We show that the proposed attack outperforms competitors on a diverse set of NLP problems for both computed metrics and human evaluation. Moreover, due to the usage of the fine-tuned language model, the generated adversarial examples are hard to detect, thus current models are not robust. Hence, it is difficult to defend from the proposed attack, which is not the case for other attacks.
Sergei Pavlovich Strelkov (obituary)
Памяти Сергея Павловича Стрелкова
1974
Uspehi fiziceskih nauk
Памяти Сергея Павловича Стрелкова
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2017; you can also visit the original URL.
The file type is application/pdf.
197ά г. Сентябрь Том 114, вып. 1 УСПЕХИ ФИЗИЧЕСКИХ НАУК PERSONALIA 53(92) ПАМЯТИ СЕРГЕЯ ПАВЛОВИЧА СТРЕЛКОВА 2 апреля 1974 г. скоропостижно скончался Сергей Павлович Стрелков, один из старейших профессоров физического факультета Московского государственного университета, заслуженный деятель науки и техники РСФСР. СП. Стрелков родился в 1905 г. в селе Долгоруково Пензенской губернии в семье учителей сельской школы. В 1927 г. он поступил на физико-математический факультет МГУ и прошел путь от
doi:10.3367/ufnr.0114.197409p.0169
fatcat:vzujxmwfrngile5yperfklwgba
more »
... нта до профессора, заведующего одной из крупнейших кафедр физического факультета. Являясь, наряду с А. А. Андроновым, А. А. Виттом, С. Э. Хайкиным, Г. С. Гореликом и др., одним из ярких представителей научной школы академика Л. И. Мандельштама, Сергей Павлович Стрелков был крупным специалистом в области физики колебаний. Многие задачи, которые ставились и решались в его работах, а также в работах его учеников были вызваны проблемами, возникающими при создании новой авиационной техники. Наиболее характерной особенностью всех научных исследований С. П. Стрелкова является глубокий физический анализ колебательных процессов, происходящих в рассматриваемых системах. Прекрасно владея сложными математическими методами, Сергей Павлович никогда не ограничивался формально-математической стороной дела. Его основной целью было раскрытие физического содержания исследуемых явлений. С. П. Стрелковым было опубликовано свыше пятидесяти научных трудов в различных областях физики и механики. Эти работы посвящены главным образом исследованию автоколебаний в различных механических системах. Сергей Павлович является автором прекрасных учебников по механике и теории колебаний, а также задачника по общему курсу физики. По этим учебникам учились и учатся десятки тысяч студентов университетов. За десять дней до смерти С. П. Стрелков сдал в издательство рукопись нового, значительно переработанного издания своей заслуженно популярной книги по механике. Когда Сергея Павловича спрашивали о том, что нового он внес в свою книгу, он отвечал, что старался прежде всего изложить более нагляднее, убедительнее, доступнее физические принципы механики. Первая научная работа * С. П. Стрелкова, опубликованная в 1933 г., была посвящена исследованию автоколебаний в механической системе (маятник Фроуда). В этой работе дана подробная физическая картина возбуждения автоколебаний в системе с трением скольжения, зависящим от скорости, проведен математический анализ процесса установления стационарной амплитуды, а также приведены результаты экспериментального иссследования. В следующей работе 2 С. П. Стрелковым дан подробный качественный и количественный анализ переходных процессов в ламповом генераторе с двумя колебательными контурами (явление затягивания частоты). Впоследствии это исследование вошло во все учебники по теории нелинейных колебаний. Сергей Павлович Стрелков был одним из основоположников исследований автоколебаний в распределенных системах. Им была полностью решена задача об автоколебаниях в двухпроводной линии с электронной лампой на одном из концов 3 . В своих работах Сергей Павлович успешно использовал строгие нелинейные методы, развиваемые школой Л. И. Мандельштама. Эти методы сущестенно отличались от применявшихся тогда в технике квазилинейных методов анализа автоколебательных систем. В то время казалось, что эти два подхода принципиально исключают друг друга, но дальнейшее развитие обоих направлений показало, что они дополняют друг друга. Большой цикл работ С. П. Стрелкова 4 ~6 посвящен исследованию автоколебаний различных тел в потоке воздуха или жидкости. Сергей Павлович, по-видимому, впер-© Издательство «Наука» \1 УФН, т. 114, вып. 1 «Успехи физических наук», 1974 г.
Shared bioinformatics databases within the Unipro UGENE platform
2015
Journal of Integrative Bioinformatics
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2020; you can also visit the original URL.
The file type is application/pdf.
Unipro UGENE is an open-source bioinformatics toolkit that integrates popular tools along with original instruments for molecular biologists within a unified user interface. Nowadays, most bioinformatics desktop applications, including UGENE, make use of a local data model while processing different types of data. Such an approach causes an inconvenience for scientists working cooperatively and relying on the same data. This refers to the need of making multiple copies of certain files for
doi:10.2390/biecoll-jib-2015-257
pmid:26527191
fatcat:sabnkr2g2vhahcfp3dnw5pw6pe
more »
... workplace and maintaining synchronization between them in case of modifications. Therefore, we focused on delivering a collaborative work into the UGENE user experience. Currently, several UGENE installations can be connected to a designated shared database and users can interact with it simultaneously. Such databases can be created by UGENE users and be used at their discretion. Objects of each data type, supported by UGENE such as sequences, annotations, multiple alignments, etc., can now be easily imported from or exported to a remote storage. One of the main advantages of this system, compared to existing ones, is the almost simultaneous access of client applications to shared data regardless of their volume. Moreover, the system is capable of storing millions of objects. The storage itself is a regular database server so even an inexpert user is able to deploy it. Thus, UGENE may provide access to shared data for users located, for example, in the same laboratory or institution. UGENE is available at: http://ugene.net/download.html.
Shared Bioinformatics Databases within the Unipro UGENE Platform
2015
Journal of Integrative Bioinformatics
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2020; you can also visit the original URL.
The file type is application/pdf.
SummaryUnipro UGENE is an open-source bioinformatics toolkit that integrates popular tools along with original instruments for molecular biologists within a unified user interface. Nowadays, most bioinformatics desktop applications, including UGENE, make use of a local data model while processing different types of data. Such an approach causes an inconvenience for scientists working cooperatively and relying on the same data. This refers to the need of making multiple copies of certain files
doi:10.1515/jib-2015-257
fatcat:pdmvoq7aofe2dbdf5ffvafbopq
more »
... r every workplace and maintaining synchronization between them in case of modifications. Therefore, we focused on delivering a collaborative work into the UGENE user experience. Currently, several UGENE installations can be connected to a designated shared database and users can interact with it simultaneously. Such databases can be created by UGENE users and be used at their discretion. Objects of each data type, supported by UGENE such as sequences, annotations, multiple alignments, etc., can now be easily imported from or exported to a remote storage. One of the main advantages of this system, compared to existing ones, is the almost simultaneous access of client applications to shared data regardless of their volume. Moreover, the system is capable of storing millions of objects. The storage itself is a regular database server so even an inexpert user is able to deploy it. Thus, UGENE may provide access to shared data for users located, for example, in the same laboratory or institution. UGENE is available at: http://ugene.net/download.html.
Multidrug-Resistant Klebsiella pneumoniae Causing Severe Infections in the Neuro-ICU
2021
Antibiotics
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2021; you can also visit the original URL.
The file type is application/pdf.
The purpose of this study was the identification of genetic lineages and antimicrobial resistance (AMR) and virulence genes in Klebsiella pneumoniae isolates associated with severe infections in the neuro-ICU. Susceptibility to antimicrobials was determined using the Vitek-2 instrument. AMR and virulence genes, sequence types (STs), and capsular types were identified by PCR. Whole-genome sequencing was conducted on the Illumina MiSeq platform. It was shown that K. pneumoniae isolates of ST14K2,
doi:10.3390/antibiotics10080979
fatcat:ef36q4zpnfaofcrcifebmdqnem
more »
... ST23K57, ST39K23, ST76K23, ST86K2, ST218K57, ST219KL125/114, ST268K20, and ST2674K47 caused severe systemic infections, including ST14K2, ST39K23, and ST268K20 that were associated with fatal incomes. Moreover, eight isolates of ST395K2 and ST307KL102/149/155 were associated with manifestations of vasculitis and microcirculation disorders. Another 12 K. pneumoniae isolates of ST395K2,KL39, ST307KL102/149/155, and ST147K14/64 were collected from patients without severe systemic infections. Major isolates (n = 38) were XDR and MDR. Beta-lactamase genes were identified: blaSHV (n = 41), blaCTX-M (n = 28), blaTEM (n = 21), blaOXA-48 (n = 21), blaNDM (n = 1), and blaKPC (n = 1). The prevalent virulence genes were wabG (n = 41), fimH (n = 41), allS (n = 41), and uge (n = 34), and rarer, detected only in the genomes of the isolates causing severe systemic infections—rmpA (n = 8), kfu (n = 6), iroN (n = 5), and iroD (n = 5) indicating high potential of the isolates for hypervirulence.
Rem Viktorovich Khokhlov (Obituary)
Памяти Рема Викторовича Хохлова
1978
Uspehi fiziceskih nauk
Памяти Рема Викторовича Хохлова
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2017; you can also visit the original URL.
The file type is application/pdf.
QUASIOPTICAL SOLID-STATE GENERATOR WITH OPEN QUASI-FREQUENCY SPHEREOSELETAL OSCILLATING SYSTEM
2021
Bulletin of the National Technical University «KhPI» Series New solutions in modern technologies
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2021; you can also visit the original URL.
The file type is application/pdf.
The results of research related to the development of an effective high-Q open oscillatory system of quasi-optical solid-state generators of millimeter and submillimeter wavelength ranges are presented. It is shown that for the stable operation of solid-state generators based on Gunn diodes and avalanche-transit diodes (ATDs) located inside an open resonator (OR), a significant rarefaction of the resonance frequency spectrum is required, which excludes jumps of the generation frequency. Taking
doi:10.20998/2413-4295.2021.02.05
fatcat:aqiufurktjbrpfgxzj452fsoie
more »
... his requirement into account, a spheroechelett open resonator (SE OR) is proposed as an open oscillatory system of solid-state generators, one of the mirrors of which is a reflective lattice of the "echelette" type. The paper presents the results of "cold" measurements of the spectrum of resonance frequencies of the SE OP, which indicate that the SE OP indeed has a significantly rarefied spectrum both in transverse and, which is very important, in longitudinal modes of oscillations. The conditions necessary for the excitation of high-quality stable oscillations with high energy efficiency in a solid-state generator based on an SE OR are discussed and implemented. The results of an experimental study of a quasi-optical generator based on an 8-mm ATD wavelength range are presented. The analysis of the obtained characteristics and their qualitative explanation on the basis of the proposed model of the processes occurring in the investigated generator are presented. It is shown that the SE OR can be used as a resonant system in the entire EHF range of not only solid-state, but also electronic-vacuum devices. Finally, a way is outlined in the direction of further rarefaction of the spectrum of the SE OR in order to create a really single-frequency open oscillatory system.
Solid-state generator of EHF range with increased long-term frequency stability
2019
Bulletin of the National Technical University «KhPI» Series New solutions in modern technologies
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2019; you can also visit the original URL.
The file type is application/pdf.
Ivan Olkhovskiy -lecturer, Kharkiv College of State University of Telecommunications; Kharkiv, Ukraine; e-mail: olkhovskiy_i_p@ukr.net. ...
Anatoliy Fursov -PhD, lecturer, Kharkiv patent-computer college; Kharkiv, Ukraine; e-mail: olkhovskiy_i_p@ukr.net. ...
doi:10.20998/2413-4295.2019.02.06
fatcat:lx5yqmqvvnganh274dmdyyb7le
Gradient-based adversarial attacks on categorical sequence models via traversing an embedded world
[article]
2020
arXiv
pre-print
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2020; you can also visit the original URL.
The file type is application/pdf.
Other Versions
Deep learning models suffer from a phenomenon called adversarial attacks: we can apply minor changes to the model input to fool a classifier for a particular example. The literature mostly considers adversarial attacks on models with images and other structured inputs. However, the adversarial attacks for categorical sequences can also be harmful. Successful attacks for inputs in the form of categorical sequences should address the following challenges: (1) non-differentiability of the target
arXiv:2003.04173v3
fatcat:4zds5zbxzzaqthax5f7quvaazi
more »
... nction, (2) constraints on transformations of initial sequences, and (3) diversity of possible problems. We handle these challenges using two black-box adversarial attacks. The first approach adopts a Monte-Carlo method and allows usage in any scenario, the second approach uses a continuous relaxation of models and target metrics, and thus allows usage of state-of-the-art methods for adversarial attacks with little additional effort. Results for money transactions, medical fraud, and NLP datasets suggest that proposed methods generate reasonable adversarial sequences that are close to original ones but fool machine learning models.
Early Response of Antimicrobial Resistance and Virulence Genes Expression in Classical, Hypervirulent, and Hybrid hvKp-MDR Klebsiella pneumoniae on Antimicrobial Stress
2021
Antibiotics
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2022; you can also visit the original URL.
The file type is application/pdf.
Klebsiella pneumoniae is an increasingly important hospital pathogen. Classical K. pneumoniae (cKp) and hypervirulent K. pneumoniae (hvKp) are two distinct evolutionary genetic lines. The recently ongoing evolution of K. pneumoniae resulted in the generation of hybrid hvKP-MDR strains. K. pneumoniae distinct isolates (n = 70) belonged to 20 sequence types with the prevalence of ST395 (27.1%), ST23 (18.6%), ST147 (15.7%), and ST86 (7.1%), and 17 capsular types with the predominance of K2
doi:10.3390/antibiotics11010007
pmid:35052884
pmcid:PMC8773033
fatcat:fiugbeylxna37gf4cuc7vczzii
more »
... K57 (18.6%), K64 (10.0%), K1 (5.7%) were isolated from patients of the Moscow neurosurgery ICU in 2014–2019. The rate of multi-drug resistant (MDR) and carbapenem-resistant phenotypes were 84.3% and 45.7%, respectively. Whole-genome sequencing of five selected strains belonging to cKp (ST395K47 and ST147K64), hvKp (ST86K2), and hvKp-MDR (ST23K1 and ST23K57) revealed blaSHV, blaTEM, blaCTX, blaOXA-48, and blaNDM beta-lactamase genes; acr, oqx, kpn, kde, and kex efflux genes; and K. pneumoniae virulence genes. Selective pressure of 100 mg/L ampicillin or 10 mg/L ceftriaxone induced changes of expression levels for named genes in the strains belonging to cKp, hvKp, and hybrid hvKp-MDR. Obtained results seem to be important for epidemiologists and clinicians for enhancing knowledge about hospital pathogens.
Genotyping, Assessment of Virulence and Antibacterial Resistance of the Rostov Strain of Mycobacterium tuberculosis Attributed to the Central Asia Outbreak Clade
2020
Pathogens
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2020; you can also visit the original URL.
The file type is application/pdf.
The Central Asia Outbreak (CAO) clade is a growing public health problem for Central Asian countries. Members of the clade belong to the narrow branch of the Mycobacterium tuberculosis Beijing genotype and are characterized by multidrug resistance and increased transmissibility. The Rostov strain of M. tuberculosis isolated in Russia and attributed to the CAO clade based on PCR-assay and whole genome sequencing and the laboratory strain H37Rv were selected to evaluate the virulence on C57Bl/6
doi:10.3390/pathogens9050335
pmid:32365818
fatcat:agocbyy6kngrtbmapbuodsine4
more »
... ce models by intravenous injection. All mice infected with the Rostov strain succumbed to death within a 48-day period, while more than half of the mice infected by the H37Rv strain survived within a 90-day period. Mice weight analysis revealed irreversible and severe depletion of animals infected with the Rostov strain compared to H37Rv. The histological investigation of lung and liver tissues of mice on the 30th day after injection of mycobacterial bacilli showed that the pattern of pathological changes generated by two strains were different. Moreover, bacterial load in the liver and lungs was higher for the Rostov strain infection. In conclusion, our data demonstrate that the drug-resistant Rostov strain exhibits a highly virulent phenotype which can be partly explained by the CAO-specific mutations.
MDR and Pre-XDR Clinical Mycobacterium tuberculosis Beijing Strains: Assessment of Virulence and Host Cytokine Response in Mice Infectious Model
2021
Microorganisms
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2021; you can also visit the original URL.
The file type is application/pdf.
Mycobacterium tuberculosis Beijing genotype associated with drug resistance is a growing public health problem worldwide. The aim of this study was the assessment of virulence for C57BL/6 mice after infection by clinical M. tuberculosis strains 267/47 and 120/26, which belong to the modern sublineages B0/W148 and Central Asia outbreak of the Beijing genotype, respectively. The sublineages were identified by the analysis of the strains' whole-genomes. The strains 267/47 and 120/26 were
doi:10.3390/microorganisms9081792
pmid:34442871
pmcid:PMC8400193
fatcat:kbqovwp4bndafchzshpxhhpgt4
more »
... zed as agents of pre-extensively drug-resistant (pre-XDR) and multidrug-resistant (MDR) tuberculosis, respectively. Both clinical strains were slow-growing in 7H9 broth compared to the M. tuberculosis H37Rv strain. The survival rates of C57BL/6 mice infected by 267/47, 120/26, and H37Rv on the 150th day postinfection were 10%, 40%, and 70%, respectively. Mycobacterial load in the lungs, spleen, and liver was higher and histopathological changes were more expressed for mice infected by the 267/47 strain compared to those infected by the 120/26 and H37Rv strains. The cytokine response in the lungs of C57BL/6 mice after infection with the 267/47, 120/26, and H37Rv strains was different. Notably, proinflammatory cytokine genes Il-1α, Il-6, Il-7, and Il-17, as well as anti-inflammatory genes Il-6 and Il-13, were downregulated after an infection caused by the 267/47 strain compared to those after infection with the H37Rv strain.
On Complexity of Searching a Subset of Vectors with Shortest Average Under a Cardinality Restriction
[chapter]
2017
Communications in Computer and Information Science
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2019; you can also visit the original URL.
The file type is application/pdf.
A Project Teams Creation based on Communities Detection 16:10 Alina Kadyrova and Konstantin Fursov Evolution of advanced technology studies: a search for communication core 16:20 ...
Networking Site Profiles
16:00
Mikhail Semenov,
Elena Koroleva,
Dilmurat Tursunov,
Lev Bulygin
A Project Teams Creation based on Communities Detection
16:10 Alina Kadyrova and
Konstantin Fursov ...
Bolsunovskaya
The 360° Around View System for Large Vehicles, the Methods of
Calibration and Removal of Barrel Distortion for Omnidirectional
Cameras
11:20
Anton Korsakov,
Ivan Fomin,
Dmitry ...
doi:10.1007/978-3-319-52920-2_5
fatcat:tbsy74foabfybosqzxmdwxryee
Opriçnina Teşkilatı ve Rus Merkezileşmesine Etkileri
2018
VAKANÜVİS - ULUSLARARASI TARİH ARAŞTIRMALARI DERGİSİ
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2020; you can also visit the original URL.
The file type is application/pdf.
Rus Devleti, XVI. yy da yaşadığı dönüşüm ve geçiş sürecine bağlı olarak iç ve dış politikada tüm tehditlere karşı İvan Groznıy kimliğinin izlerini taşıyan bir teşkilatlanma oluşturmuştur. ...
Andrey Fursov bu yaklaşımın Altınordu fetih politikalarında uyguladığı toprak sistemi ile aynı olduğunu belirtmektedir. Bknz: Fursov, a.g.m. s. 150. 60 Froyanov, a.g.e., s.5 61 İ. A. ...
A.L Yurganov, V.A Şarov, D.N Alışin, Y.V Anhimuk ve Andrey Fursov gibi Rus araştırmacılar Opriçnina hakkında önemli çalışmalar yapmışlardır. ...
doaj:2280f926ed934afebf4b4cb3603c3b59
fatcat:g5aj7pxwdzc2rhgno2zcprn2n4
« Previous
Showing results 1 — 15 out of 43 results