BEGIN:VCALENDAR
PRODID:-//Columba Systems Ltd//NONSGML CPNG/SpringViewer/ICal Output/3.3-
 M3//EN
VERSION:2.0
CALSCALE:GREGORIAN
METHOD:PUBLISH
BEGIN:VEVENT
DTSTAMP:20220107T130516Z
DTSTART:20220216T140000Z
DTEND:20220216T150000Z
SUMMARY:Ritabrata Dutta  - Probabilistic Forecasting with Conditional Gen
 erative Networks via Scoring Rule Minimization
UID:{http://www.columbasystems.com/customers/uom/gpp/eventid/}ql3-ku8g0lu
 d-von4v2
DESCRIPTION:Ritabrata Dutta\, Assistant Professor of Statistics at the De
 partment of Statistics\, University of Warwick is our speaker for the St
 atistics seminar series.\n\nTitle: Probabilistic Forecasting with Condit
 ional Generative Networks via Scoring Rule Minimization\n\nAbstract: Pro
 babilistic forecasting consists of stating a probability distribution fo
 r a future outcome based on past observations. In meteorology\, ensemble
 s of physics-based numerical models are run to get such distribution. Us
 ually\, performance is evaluated with scoring rules\, functions of the f
 orecast distribution and the observed outcome. With some scoring rules\,
  calibration and sharpness of the forecast can be assessed at the same t
 ime. \nIn deep learning\, generative neural networks parametrize distrib
 utions on high-dimensional spaces and easily allow sampling by transform
 ing draws from a latent variable. Conditional generative networks additi
 onally constrain the distribution on an input variable. In this manuscri
 pt\, we perform probabilistic forecasting with conditional generative ne
 tworks trained to minimize scoring rule values. In contrast to Generativ
 e Adversarial Networks (GANs)\, no discriminator is required and trainin
 g is stable. We perform experiments on two chaotic models and a global d
 ataset of weather observations\; results are satisfactory and better cal
 ibrated than what achieved by GANs.
STATUS:TENTATIVE
TRANSP:TRANSPARENT
CLASS:PUBLIC
LOCATION:Zoom link: https://zoom.us/j/92947173491
END:VEVENT
END:VCALENDAR