BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//jEvents 2.0 for Joomla//EN
CALSCALE:GREGORIAN
METHOD:PUBLISH
BEGIN:VEVENT
UID:cor3id1j60rmab9h65hjcb9k6hhj0b9ocgp36bb171hm2dr168ojgp9hck@google.com
CATEGORIES:RTportal.ru
CREATED:20170606T092329
SUMMARY:семинар "Структурные модели и глубинное обучение": дополнительный семинар
LOCATION:ИППИ РАН, аудитория 615
DESCRIPTION;ENCODING=QUOTED-PRINTABLE:Коллеги!<br/><br/>На дополнительном заседании семинара "Структурные модели 
 и глубинное обучение", которое состоится 17 июня (пятница), в 16.00, ИППИ Р
 АН (<a href="http://iitp.ru/ru/contacts.htm">http://iitp.ru/ru/contacts.htm
 </a>), 6 этаж, 615 аудитория<br/><br/>будет представлено несколько коротких
  сообщений о различных интересных направлениях машинного обучения.<br/><br/
 >===================<br/><br/>Author: Denis Volhonskiy (HSE, IITP)<br/><br/
 >Title: Deep Convolutional Generative Adversarial Networks in Steganography
 <br/><br/>Annotation: Steganography is the way of hiding information within
  other information (called container). Steganalysis is the study of detecti
 ng messages hidden using steganography --- usually it is binary classifier 
 (detect if there is some information / no information in container).<br/><b
 r/>We propose a new model based on Deep Convolutional Generative Adversaria
 l Networks for generating images, that could be used for more safety inform
 ation embedding (in terms of steganalysis accuracy) using steganography alg
 orithms. Our model allows increasing steganalysis error on a test set of ge
 nerated images in comparison with an original test set.<br/><br/>==========
 =========<br/><br/>Author: Vladislav Ishimtsev (HSE, IITP)<br/><br/>Title: 
 Conformalized density- and distance-based anomaly detection in time-series 
 data<br/><br/>Annotation: Most of the world's data is streaming, time serie
 s data, where anomalies give pertinent information in critical situations; 
 Examples abound in such fields as finance, IT, security, health and energy.
  However, detection of anomalies in time-series data is a difficult task, r
 equiring detectors to process data in real-time, not batches, and learn whi
 le simultaneously making predictions. <br/><br/>We consider new approaches 
 to detect anomalies in time-series data using conformalized density- and di
 stance-based anomaly detection algorithms. <br/><br/>Testing and comparison
  of algorithms will be done on Numenta Anomaly Benchmark (NAB), which attem
 pts to provide a controlled and repeatable environment of open-source tools
  to test and measure anomaly detection algorithms on streaming data.<br/><b
 r/>===================<br/><br/>Author: Albert Matveev (HSE)<br/><br/>Title
 : Experts Aggregation for Time Series Prediction<br/><br/>Annotation: Time 
 series prediction is one of the most significant problems in applied mathem
 atics. Estimation of a future outcome of some sequence is desired in many s
 cientific and practical fields, especially in finance. Regression-type fore
 casters are usually used in order to solve this problem. However, it is ess
 ential to note that for different regimes we need to use different methods.
  It is obvious that without knowing true future value of a time series, the
  learner cannot select a method and corresponding forecast, which would pro
 vide the best accuracy among available ones. Therefore, we consider a set o
 f independent base forecasting algorithms which we call experts and formula
 te a problem of prediction with expert advice. We would like to construct a
 n aggregating algorithm that will adaptively aggregate available expert for
 ecasts so that aggregated forecast is close to the best expert in terms of 
 loss process. <br/><br/>In this talk the main aggregating schemes and loss 
 bounds for them will be presented. These aggregating algorithms will be com
 pared on several datasets and the results of computational experiments will
  be provided.<br/><br/>===================<br/><br/>Author: Oleg Maslenniko
 v (HSE)<br/><br/>Title: Deep Features based Image Retrieval<br/><br/>Annota
 tion: Computer vision tasks are common in today's world, for example, OCR t
 asks (optical character recognition), license plate recognition with traffi
 c cameras, image processing in medicine and others. One of these tasks is C
 BIR - content base image retrieval. Examples of such tasks are the image se
 arch in Google, recently appeared service “Findface”, which allows finding 
 a person in social networks by his/her photo.<br/><br/>In the presentation 
 we will review approaches to deep learning based solution of CBIR tasks, di
 scuss various architectures of corresponding neural networks, as well as th
 e choice of a metric distance and its influence on obtained results.<br/><b
 r/>--<br/><br/>С уважением,<br/><br/>Евгений Бурнаев This event was importe
 d from: https://rtportal.ru/index.php/kalendar-sobytij/eventdetail/136/-/se
 minar-strukturnye-modeli-i-glubinnoe-obuchenie-dopolnitelnyj-seminar?tmpl=c
 omponent
DTSTAMP:20260812T201355Z
DTSTART;TZID=Europe/Moscow:20160617T160000
DTEND;TZID=Europe/Moscow:20160617T173000
SEQUENCE:0
TRANSP:OPAQUE
END:VEVENT
END:VCALENDAR