BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Memento EPFL//
BEGIN:VEVENT
SUMMARY:Project PHI: System Design for Pervasive Hierarchal Intelligence
DTSTART:20170523T103000
DTEND:20170523T114500
DTSTAMP:20260407T102609Z
UID:952b78f64c15e164392e76acb98059aa2ad2fba9569d6e959057e360
CATEGORIES:Conferences - Seminars
DESCRIPTION:Hadi Esmaeilzadeh\, Georgia Tech's School of Computer Science\
 nThis talk presents\, Project PHI (Pervasive Hierarchical Intelligence) a 
 holistic effort to provide a comprehensive solution for making immersive m
 achine intelligence a reality.  \n\nOur guiding principle is to retain as
  much generality and automation while delivering large performance and eff
 iciency gains through specialization and acceleration for a wide range of 
 learning and intelligence workloads. As the first milestones of Project PH
 I\, we have developed Tabla and DnnWeaver\, which are open source and publ
 ically available (http://act-lab.org/artifacts/tabla/ and http://act-lab.o
 rg/artifacts/dnnweaver/). DnnWeaver is the very first open-source hardware
  acceleration framework for deep neural networks. Tabla is a cross-stack s
 olution—spanning from programming language to the hardware—that rethin
 ks the hardware/software abstraction by delving into the theory of machine
  learning. It leverages the insight that many learning algorithms can be s
 olved using stochastic gradient descent that minimizes an objective functi
 on. The solver is fixed while the objective function changes with the lear
 ning algorithm. Therefore\, Tabla uses stochastic optimization as the abst
 raction between hardware and software.\n\nConsequently\, programmers speci
 fy the learning algorithm by merely expressing the gradient of the objecti
 ve function in our domain specific language. Tabla then automatically gene
 rates the synthesizable implementation of the accelerator for scale-out FP
 GA realization using a set of template designs. Real hardware measurements
  show orders of magnitude higher performance and power efficiency while th
 e programmer only writes 60 lines of code. These encouraging results show 
 that rethinking the hardware/software abstractions from an algorithmic per
 spective can open new dimensions in system design for Pervasive Hierarchic
 al Intelligence.
LOCATION:BC 420 https://plan.epfl.ch/?room==BC%20420
STATUS:CONFIRMED
END:VEVENT
END:VCALENDAR