BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Adversarial generation of gene expression data -  Ramon Vinas Torn
 e\, UCL
DTSTART:20181122T170000Z
DTEND:20181122T180000Z
UID:TALK116233@talks.cam.ac.uk
CONTACT:Pietro Lio
DESCRIPTION:The problem of reverse engineering gene regulatory networks fr
 om\nhigh-throughput expression data is one of the biggest challenges in\nb
 ioinformatics. In order to benchmark network inference algorithms\, simula
 tors\nof well-characterized expression datasets are often required. Howeve
 r\, existing\nsimulators have been criticized because they fail to emulate
  key properties of\ngene expression data (Maier et al.\, 2013).\n\nThe pur
 pose of this work is two-fold. First\, we study and propose mechanisms to\
 nfaithfully assess the realism of a synthetic expression dataset. Second\,
  we\ndesign an adversarial simulator of expression data\, gGAN\, based on 
 a generative\nadversarial network (Goodfellow et al.\, 2014). We show that
  our model\noutperforms existing simulators by a large margin in terms of 
 the realism of\nthe generated data. More importantly\, our results show th
 at gGAN is\, to our\nbest knowledge\, the first simulator that passes the 
 Turing test for gene\nexpression data proposed by Maier et al. (2013).
LOCATION:Department of Computer Science and technology\, sw01 
END:VEVENT
END:VCALENDAR