BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Memento EPFL//
BEGIN:VEVENT
SUMMARY:candidacy exam
DTSTART:20180301T140000
DTEND:20180301T160000
DTSTAMP:20260407T034509Z
UID:f6ef8a6a1886c83defc081f3f5123610eed38cb5549ae04e8bd09899
CATEGORIES:Conferences - Seminars
DESCRIPTION:Rabeeh Karimi\nEDIC candidacy exam\nExam president: Prof. Pasc
 al Frossard\nThesis advisor: Prof. Volkan Cevher\nCo-examiner: Prof. Alexa
 ndre Alahi\n\nAbstract\nAbstract—Nowadays deep learning methods have ach
 ieved\nstate-of-the-art performance in many fields\, including signal\npro
 cessing. Such data-driven methods are learning the signal\nstructure from 
 large training data\, neglecting conventional signal\nrecovery approaches.
  It is worth investigating how infusing\nconventional signal processing me
 thods into deep learning models\ncould benefit current models with the hop
 e to solve a wider range\nof problems\, and increasing the performance and
  accuracy of\ndeep learning architectures. Towards this goal\, we explain 
 briefly\nhow to jointly train sampling pattern and reconstruction network\
 nfor signal recovery task given compressive measurements.\n\nBackground pa
 pers\nLearning-Based Compressive Subsampling\, by Luca Baldassarre et all
 .\nLearning to invert: Signal recovery via deep convolutional networks\, b
 y Ali. Mousavi and Richard G.Baraniuk \nhttps://arxiv.org/pdf/1701.03891.
 pdf\nLearning representations from EGG with deep recurrent convolutional n
 eural networks\, by \nPouya Bashivan et all.\n 
LOCATION:ELD 120 https://plan.epfl.ch/?room=ELD120
STATUS:CONFIRMED
END:VEVENT
END:VCALENDAR