-
AutoFraudNet: A Multimodal Network to Detect Fraud in the Auto Insurance Industry
Authors:
Azin Asgarian,
Rohit Saha,
Daniel Jakubovitz,
Julia Peyre
Abstract:
In the insurance industry detecting fraudulent claims is a critical task with a significant financial impact. A common strategy to identify fraudulent claims is looking for inconsistencies in the supporting evidence. However, this is a laborious and cognitively heavy task for human experts as insurance claims typically come with a plethora of data from different modalities (e.g. images, text and m…
▽ More
In the insurance industry detecting fraudulent claims is a critical task with a significant financial impact. A common strategy to identify fraudulent claims is looking for inconsistencies in the supporting evidence. However, this is a laborious and cognitively heavy task for human experts as insurance claims typically come with a plethora of data from different modalities (e.g. images, text and metadata). To overcome this challenge, the research community has focused on multimodal machine learning frameworks that can efficiently reason through multiple data sources. Despite recent advances in multimodal learning, these frameworks still suffer from (i) challenges of joint-training caused by the different characteristics of different modalities and (ii) overfitting tendencies due to high model complexity. In this work, we address these challenges by introducing a multimodal reasoning framework, AutoFraudNet (Automobile Insurance Fraud Detection Network), for detecting fraudulent auto-insurance claims. AutoFraudNet utilizes a cascaded slow fusion framework and state-of-the-art fusion block, BLOCK Tucker, to alleviate the challenges of joint-training. Furthermore, it incorporates a light-weight architectural design along with additional losses to prevent overfitting. Through extensive experiments conducted on a real-world dataset, we demonstrate: (i) the merits of multimodal approaches, when compared to unimodal and bimodal methods, and (ii) the effectiveness of AutoFraudNet in fusing various modalities to boost performance (over 3\% in PR AUC).
△ Less
Submitted 15 January, 2023;
originally announced January 2023.
-
Development of an advanced Compton telescope for MeV-range gamma-ray astronomy
Authors:
Adrien Laviron,
Valentin Gourlaouen,
Clarisse Hamadache,
Corentin Hiver,
Jürgen Kiener,
Jean Peyré,
Vincent Tatischeff
Abstract:
An advanced Compton telescope appears to be the best instrument concept for the next generation gamma-ray space observatory in the MeV range. A first prototype of advanced Compton telescope is being developed to match the constraints of a nano satellite mission, with the scientific objective of measuring gamma-ray burst prompt emission polarization. Our instrumental developments for this project a…
▽ More
An advanced Compton telescope appears to be the best instrument concept for the next generation gamma-ray space observatory in the MeV range. A first prototype of advanced Compton telescope is being developed to match the constraints of a nano satellite mission, with the scientific objective of measuring gamma-ray burst prompt emission polarization. Our instrumental developments for this project are focusing on the position-sensitive calorimeter module, made of a monolithic inorganic CeBr$_3$ scintillator read by a pixelated photodetector. 3D position reconstruction is obtained by deep-learning algorithms that have been optimized down to an uncertainty of 2 mm for each spatial direction.
△ Less
Submitted 26 February, 2020;
originally announced February 2020.
-
Gamma-ray emission in alpha-particle reactions with C, Mg, Si, Fe
Authors:
J. Kiener,
J. Bundesmann,
I. Deloncle,
A. Denker,
V. Tatischeff,
A. Gostojic,
C. Hamadache,
J. Röhrich,
H. Benhabiles,
I. Bougaoub,
A. Coc,
F. Hammache,
J. Peyré
Abstract:
Cross sections for the strongest gamma-ray emission lines produced in alpha-particle reactions with C, Mg, Si, Fe have been measured in the range E_alpha = 50 - 90 MeV at the center for proton therapy at the Helmholtz-Zentrum Berlin. Data for more than 60 different gamma-ray lines were determined, with particular efforts for lines that are in cross section compilations/evaluations with astrophysic…
▽ More
Cross sections for the strongest gamma-ray emission lines produced in alpha-particle reactions with C, Mg, Si, Fe have been measured in the range E_alpha = 50 - 90 MeV at the center for proton therapy at the Helmholtz-Zentrum Berlin. Data for more than 60 different gamma-ray lines were determined, with particular efforts for lines that are in cross section compilations/evaluations with astrophysical purpose, and where data exist at lower projectile energies. The data are compared with predictions of a modern nuclear reaction code and cross-section curves of the latest evaluation for gamma-ray line emission in accelerated-particle interactions in solar flares.
△ Less
Submitted 10 January, 2020;
originally announced January 2020.
-
Detecting unseen visual relations using analogies
Authors:
Julia Peyre,
Ivan Laptev,
Cordelia Schmid,
Josef Sivic
Abstract:
We seek to detect visual relations in images of the form of triplets t = (subject, predicate, object), such as "person riding dog", where training examples of the individual entities are available but their combinations are unseen at training. This is an important set-up due to the combinatorial nature of visual relations : collecting sufficient training data for all possible triplets would be ver…
▽ More
We seek to detect visual relations in images of the form of triplets t = (subject, predicate, object), such as "person riding dog", where training examples of the individual entities are available but their combinations are unseen at training. This is an important set-up due to the combinatorial nature of visual relations : collecting sufficient training data for all possible triplets would be very hard. The contributions of this work are three-fold. First, we learn a representation of visual relations that combines (i) individual embeddings for subject, object and predicate together with (ii) a visual phrase embedding that represents the relation triplet. Second, we learn how to transfer visual phrase embeddings from existing training triplets to unseen test triplets using analogies between relations that involve similar objects. Third, we demonstrate the benefits of our approach on three challenging datasets : on HICO-DET, our model achieves significant improvement over a strong baseline for both frequent and unseen triplets, and we observe similar improvement for the retrieval of unseen triplets with out-of-vocabulary predicates on the COCO-a dataset as well as the challenging unusual triplets in the UnRel dataset.
△ Less
Submitted 22 September, 2019; v1 submitted 13 December, 2018;
originally announced December 2018.
-
The e-ASTROGAM gamma-ray space observatory for the multimessenger astronomy of the 2030s
Authors:
V. Tatischeff,
A. De Angelis,
M. Tavani,
I. Grenier,
U. Oberlack,
L. Hanlon,
R. Walter,
A. Argan,
P. von Ballmoos,
A. Bulgarelli,
I. Donnarumma,
M. Hernanz,
I. Kuvvetli,
M. Mallamaci,
M. Pearce,
A. Zdziarski,
A. Aboudan,
M. Ajello,
G. Ambrosi,
D. Bernard,
E. Bernardini,
V. Bonvicini,
A. Brogna,
M. Branchesi,
C. Budtz-Jorgensen
, et al. (52 additional authors not shown)
Abstract:
e-ASTROGAM is a concept for a breakthrough observatory space mission carrying a gamma-ray telescope dedicated to the study of the non-thermal Universe in the photon energy range from 0.15 MeV to 3 GeV. The lower energy limit can be pushed down to energies as low as 30 keV for gamma-ray burst detection with the calorimeter. The mission is based on an advanced space-proven detector technology, with…
▽ More
e-ASTROGAM is a concept for a breakthrough observatory space mission carrying a gamma-ray telescope dedicated to the study of the non-thermal Universe in the photon energy range from 0.15 MeV to 3 GeV. The lower energy limit can be pushed down to energies as low as 30 keV for gamma-ray burst detection with the calorimeter. The mission is based on an advanced space-proven detector technology, with unprecedented sensitivity, angular and energy resolution, combined with remarkable polarimetric capability. Thanks to its performance in the MeV-GeV domain, substantially improving its predecessors, e-ASTROGAM will open a new window on the non-thermal Universe, making pioneering observations of the most powerful Galactic and extragalactic sources, elucidating the nature of their relativistic outflows and their effects on the surroundings. With a line sensitivity in the MeV energy range one to two orders of magnitude better than previous and current generation instruments, e-ASTROGAM will determine the origin of key isotopes fundamental for the understanding of supernova explosion and the chemical evolution of our Galaxy. The mission will be a major player of the multiwavelength, multimessenger time-domain astronomy of the 2030s, and provide unique data of significant interest to a broad astronomical community, complementary to powerful observatories such as LISA, LIGO, Virgo, KAGRA, the Einstein Telescope and the Cosmic Explorer, IceCube-Gen2 and KM3NeT, SKA, ALMA, JWST, E-ELT, LSST, Athena, and the Cherenkov Telescope Array.
△ Less
Submitted 13 July, 2018; v1 submitted 16 May, 2018;
originally announced May 2018.
-
Weakly-supervised learning of visual relations
Authors:
Julia Peyre,
Ivan Laptev,
Cordelia Schmid,
Josef Sivic
Abstract:
This paper introduces a novel approach for modeling visual relations between pairs of objects. We call relation a triplet of the form (subject, predicate, object) where the predicate is typically a preposition (eg. 'under', 'in front of') or a verb ('hold', 'ride') that links a pair of objects (subject, object). Learning such relations is challenging as the objects have different spatial configura…
▽ More
This paper introduces a novel approach for modeling visual relations between pairs of objects. We call relation a triplet of the form (subject, predicate, object) where the predicate is typically a preposition (eg. 'under', 'in front of') or a verb ('hold', 'ride') that links a pair of objects (subject, object). Learning such relations is challenging as the objects have different spatial configurations and appearances depending on the relation in which they occur. Another major challenge comes from the difficulty to get annotations, especially at box-level, for all possible triplets, which makes both learning and evaluation difficult. The contributions of this paper are threefold. First, we design strong yet flexible visual features that encode the appearance and spatial configuration for pairs of objects. Second, we propose a weakly-supervised discriminative clustering model to learn relations from image-level labels only. Third we introduce a new challenging dataset of unusual relations (UnRel) together with an exhaustive annotation, that enables accurate evaluation of visual relation retrieval. We show experimentally that our model results in state-of-the-art results on the visual relationship dataset significantly improving performance on previously unseen relations (zero-shot learning), and confirm this observation on our newly introduced UnRel dataset.
△ Less
Submitted 29 July, 2017;
originally announced July 2017.
-
The e-ASTROGAM mission (exploring the extreme Universe with gamma rays in the MeV-GeV range)
Authors:
Alessandro De Angelis,
Vincent Tatischeff,
Marco Tavani,
Uwe Oberlack,
Isabelle A. Grenier,
Lorraine Hanlon,
Roland Walter,
Andrea Argan,
Peter von Ballmoos,
Andrea Bulgarelli,
Immacolata Donnarumma,
Margarita Hernanz,
Irfan Kuvvetli,
Mark Pearce,
Andrzej Zdziarski,
Alessio Aboudan,
Marco Ajello,
Giovanni Ambrosi,
Denis Bernard,
Elisa Bernardini,
Valter Bonvicini,
Andrea Brogna,
Marica Branchesi,
Carl Budtz-Jorgensen,
Andrei Bykov
, et al. (49 additional authors not shown)
Abstract:
e-ASTROGAM (`enhanced ASTROGAM') is a breakthrough Observatory mission dedicated to the study of the non-thermal Universe in the photon energy range from 0.3 MeV to 3 GeV. The mission is based on an advanced space-proven detector technology, with unprecedented sensitivity, angular and energy resolution, combined with polarimetric capability. In the largely unexplored MeV-GeV domain, e-ASTROGAM wil…
▽ More
e-ASTROGAM (`enhanced ASTROGAM') is a breakthrough Observatory mission dedicated to the study of the non-thermal Universe in the photon energy range from 0.3 MeV to 3 GeV. The mission is based on an advanced space-proven detector technology, with unprecedented sensitivity, angular and energy resolution, combined with polarimetric capability. In the largely unexplored MeV-GeV domain, e-ASTROGAM will open a new window on the non-thermal Universe, making pioneering observations of the most powerful Galactic and extragalactic sources, elucidating the nature of their relativistic outflows and their effects on Galactic ecosystems. With a line sensitivity in the MeV energy range one to two orders of magnitude better than previous generation instruments, will determine the origin of key isotopes fundamental for the understanding of supernova explosion and the chemical evolution of our Galaxy. The mission will provide unique data of significant interest to a broad astronomical community, complementary to powerful observatories such as LIGO-Virgo-GEO600-KAGRA, SKA, ALMA, E-ELT, TMT, LSST, JWST, Athena, CTA, IceCube, KM3NeT, and the promise of eLISA.
Keywords: High-energy gamma-ray astronomy, High-energy astrophysics, Nuclear Astrophysics, Compton and Pair creation telescope, Gamma-ray bursts, Active Galactic Nuclei, Jets, Outflows, Multiwavelength observations of the Universe, Counterparts of gravitational waves, Fermi, Dark Matter, Nucleosynthesis, Early Universe, Supernovae, Cosmic Rays, Cosmic antimatter.
△ Less
Submitted 4 June, 2017; v1 submitted 7 November, 2016;
originally announced November 2016.
-
The e-ASTROGAM gamma-ray space mission
Authors:
V. Tatischeff,
M. Tavani,
P. von Ballmoos,
L. Hanlon,
U. Oberlack,
A. Aboudan,
A. Argan,
D. Bernard,
A. Brogna,
A. Bulgarelli,
A. Bykov,
R. Campana,
P. Caraveo,
M. Cardillo,
P. Coppi,
A. De Angelis,
R. Diehl,
I. Donnarumma,
V. Fioretti,
A. Giuliani,
I. Grenier,
J. E. Grove,
C. Hamadache,
D. Hartmann,
M. Hernanz
, et al. (26 additional authors not shown)
Abstract:
The e-ASTROGAM is a gamma-ray space mission to be proposed as the M5 Medium-size mission of the European Space Agency. It is dedicated to the observation of the Universe with unprecedented sensitivity in the energy range 0.2 - 100 MeV, extending up to GeV energies, together with a groundbreaking polarization capability. It is designed to substantially improve the COMPTEL and Fermi sensitivities in…
▽ More
The e-ASTROGAM is a gamma-ray space mission to be proposed as the M5 Medium-size mission of the European Space Agency. It is dedicated to the observation of the Universe with unprecedented sensitivity in the energy range 0.2 - 100 MeV, extending up to GeV energies, together with a groundbreaking polarization capability. It is designed to substantially improve the COMPTEL and Fermi sensitivities in the MeV-GeV energy range and to open new windows of opportunity for astrophysical and fundamental physics space research. e-ASTROGAM will operate as an open astronomical observatory, with a core science focused on (1) the activity from extreme particle accelerators, including gamma-ray bursts and active galactic nuclei and the link of jet astrophysics to the new astronomy of gravitational waves, neutrinos, ultra-high energy cosmic rays, (2) the high-energy mysteries of the Galactic center and inner Galaxy, including the activity of the supermassive black hole, the Fermi Bubbles, the origin of the Galactic positrons, and the search for dark matter signatures in a new energy window; (3) nucleosynthesis and chemical evolution, including the life cycle of elements produced by supernovae in the Milky Way and the Local Group of galaxies. e-ASTROGAM will be ideal for the study of high-energy sources in general, including pulsars and pulsar wind nebulae, accreting neutron stars and black holes, novae, supernova remnants, and magnetars. And it will also provide important contributions to solar and terrestrial physics. The e-ASTROGAM telescope is optimized for the simultaneous detection of Compton and pair-producing gamma-ray events over a large spectral band. It is based on a very high technology readiness level for all subsystems and includes many innovative features for the detectors and associated electronics.
△ Less
Submitted 31 January, 2017; v1 submitted 12 August, 2016;
originally announced August 2016.
-
Single electron response and energy resolution of a Micromegas detector
Authors:
T. Zerguerras,
B. Genolini,
V. Lepeltier,
J. Peyre,
J. Pouthas,
P. Rosier
Abstract:
Micro-Pattern Gaseous Detectors (MPGDs) such as Micromegas or GEM are used in particle physics experiments for their capabilities in particle tracking at high rates. Their excellent position resolutions are well known but their energy characteristics have been less studied. The energy resolution is mainly affected by the ionisation processes and detector gain fluctuations. This paper presents a…
▽ More
Micro-Pattern Gaseous Detectors (MPGDs) such as Micromegas or GEM are used in particle physics experiments for their capabilities in particle tracking at high rates. Their excellent position resolutions are well known but their energy characteristics have been less studied. The energy resolution is mainly affected by the ionisation processes and detector gain fluctuations. This paper presents a method to separetely measure those two contributions to the energy resolution of a Micromegas detector. The method relies on the injection of a controlled number of electrons. The Micromegas has a 1.6-mm drift zone and a 160-$μ$m amplification gap. It is operated in Ne 95%-iC$\mathrm{_4}$H$\mathrm{_{10}}$ 5% at atmospheric pressure. The electrons are generated by non-linear photoelectric emission issued from the photons of a pulsed 337-nm wavelength laser coupled to a focusing system. The single electron response has been measured at different gains (3.7 10$\mathrm{^4}$, 5.0 10$\mathrm{^4}$ and 7.0 10$\mathrm{^4}$) and is fitted with a good agreement by a Polya distribution. From those fits, a relative gain variance of 0.31$\pm$0.02 is deduced. The setup has also been characterised at several voltages by fitting the energy resolution measured as a function of the number of primary electrons, ranging from 5 up to 210. A maximum value of the Fano factor (0.37) has been estimated for a 5.9 keV X-rays interacting in the Ne 95%-iC$\mathrm{_4}$H$\mathrm{_{10}}$ 5% gas mixture.
△ Less
Submitted 19 January, 2009;
originally announced January 2009.
-
PMm2: large photomultipliers and innovative electronics for the next-generation neutrino experiments
Authors:
B. Genolini,
P. Barrillon,
S. Blin,
J. -E. Campagne,
B. Combettes,
S. Conforti,
A. -G. De-haine,
D. Duchesneau,
F. Dulucq,
N. Dumont-Dayot,
J. Favier,
F. Fouché,
R. Hermel,
C. de La Taille,
G. Martin-Chassard,
T. Nguyen Trung,
C. Périnet,
J. Peyré,
J. Pouthas,
L. Raux,
E. Rindel,
P. Rosier,
J. Tassan-Viol,
W. Wei,
A. Zghiche
Abstract:
The next generation of proton decay and neutrino experiments, the post-SuperKamiokande detectors as those that will take place in megaton size water tanks, will require very large surfaces of photodetection and a large volume of data. Even with large hemispherical photomultiplier tubes, the expected number of channels should reach hundreds of thousands. A funded R&D program to implement a soluti…
▽ More
The next generation of proton decay and neutrino experiments, the post-SuperKamiokande detectors as those that will take place in megaton size water tanks, will require very large surfaces of photodetection and a large volume of data. Even with large hemispherical photomultiplier tubes, the expected number of channels should reach hundreds of thousands. A funded R&D program to implement a solution is presented here. The very large surface of photodetection is segmented in macro pixels made of 16 hemispherical (12 inches) photomultiplier tubes connected to an autonomous front-end which works on a triggerless data acquisition mode. The expected data transmission rate is 5 Mb/s per cable, which can be achieved with existing techniques. This architecture allows to reduce considerably the cost and facilitate the industrialization. This document presents the simulations and measurements which define the requirements for the photomultipliers and the electronics. A proto-type of front-end electronics was successfully tested with 16 photomultiplier tubes supplied by a single high voltage, validating the built-in gain adjustment and the calibration principle. The first tests and calculations on the photomultiplier glass led to the study of a new package optimized for a 10 bar pressure in order to sustain the high underwater pressure.
△ Less
Submitted 17 November, 2008;
originally announced November 2008.
