BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:America/Chicago
X-LIC-LOCATION:America/Chicago
BEGIN:DAYLIGHT
TZOFFSETFROM:-0600
TZOFFSETTO:-0500
TZNAME:CDT
DTSTART:19700308T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=2SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:-0500
TZOFFSETTO:-0600
TZNAME:CST
DTSTART:19701101T020000
RRULE:FREQ=YEARLY;BYMONTH=11;BYDAY=1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20260522T150153Z
LOCATION:D167/174
DTSTART;TZID=America/Chicago:20181112T090000
DTEND;TZID=America/Chicago:20181112T173000
UID:submissions.supercomputing.org_SC18_sess151@linklings.com
SUMMARY:Machine Learning in HPC Environments
DESCRIPTION:Automated Labeling of Electron Microscopy Images Using Deep Le
 arning\n\nSearching for scientific data requires metadata providing a rele
 vant context. Today, generating metadata is a time and labor intensive man
 ual process that is often neglected, and important datasets are not access
 ible through search. We investigate the use of machine learning to general
 ize metadata f...\n\n\nGunther Weber (Lawrence Berkeley National Laborator
 y; University of California, Davis) and Colin Ophus and Lavanya Ramakrishn
 an (Lawrence Berkeley National Laboratory)\n---------------------\nWorksho
 p Lunch (on your own)\n---------------------\nCommunication-Efficient Para
 llelization Strategy for Deep Convolutional Neural Network Training\n\nTra
 ining modern Convolutional Neural Network (CNN) models is extremely time-c
 onsuming, and the efficiency of its parallelization plays a key role in fi
 nishing the training in a reasonable amount of time. The well-known parall
 el synchronous Stochastic Gradient Descent (SGD) algorithm suffers from hi
 g...\n\n\nSunwoo Lee and Ankit Agrawal (Northwestern University), Prasanna
  Balaprakash (Argonne National Laboratory), and Alok Choudhary and Wei-ken
 g Liao (Northwestern University)\n---------------------\nWorkshop Morning 
 Break\n---------------------\nScaling Deep Learning for Cancer with Advanc
 ed Workflow Storage Integration\n\nCancer Deep Learning Environment (CANDL
 E) benchmarks and workflows will combine the power of exascale computing w
 ith neural network-based machine learning to address a range of loosely co
 nnected problems in cancer research.  This application area poses unique c
 hallenges to the exascale computing env...\n\n\nJustin Wozniak (Argonne Na
 tional Laboratory); Philip Davis (Rutgers University); Tong Shu, Jonathan 
 Ozik, and Nicholson Collier (Argonne National Laboratory); Manish Parashar
  (Rutgers University); and Ian Foster, Thomas Brettin, and Rick Stevens (A
 rgonne National Laboratory)\n---------------------\nWorkshop Overview\n---
 ------------------\nWorkshop Afternoon Break\n---------------------\nLarge
  Minibatch Training on Supercomputers with Improved Accuracy and Reduced T
 ime to Train\n\nFor the past 6 years, the ILSVRC competition and the Image
 Net dataset have attracted a lot of interest from the Computer Vision comm
 unity, allowing for state-of-the-art accuracy to grow tremendously. This s
 hould be credited to the use of deep artificial neural network designs. As
  these became more c...\n\n\nValeriu Codreanu and Damian Podareanu (SURFsa
 ra) and Vikram Saletore (Intel Corporation)\n---------------------\nIntrod
 uction - Machine Learning in HPC Environments\n\nThe intent of this worksh
 op is to bring together researchers, practitioners, and scientific communi
 ties to discuss methods that utilize extreme scale systems for machine lea
 rning. This workshop will focus on the greatest challenges in utilizing HP
 C for machine learning and methods for exploiting dat...\n\n\nSteven R. Yo
 ung and Robert M. Patton (Oak Ridge National Laboratory), Janis Keuper (Fr
 aunhofer Institute for Industrial Mathematics), and Michael Houston (Nvidi
 a Corporation)\n---------------------\nMorning Keynote – Azalia Mirhoseini
  (Google)\n\nAdvances in computer systems have been key to the success of 
 Machine Learning (ML) in recent years. With the ubiquitous success of ML, 
 it is now time for a new era where we can transform the way computer syste
 ms are built -- with learning. This talk highlights some of the challenges
  that modern comp...\n\n\nAzalia Mirhoseini (Google LLC)\n----------------
 -----\nWorkshop Overview\n---------------------\nOptimizing Machine Learni
 ng on Apache Spark in HPC Environments\n\nMachine learning has established
  itself as a powerful tool for the construction of decision making models 
 and algorithms through the use of statistical techniques on training data.
  However, a significant impediment to its progress is the time spent train
 ing and improving the accuracy of these models...\n\n\nZhenyu Li, James Da
 vis, and Stephen Jarvis (University of Warwick)\n---------------------\nLa
 rge-Scale Clustering Using MPI-Based Canopy\n\nAnalyzing massive amounts o
 f data and extracting value from it has become key across different discip
 lines. Many approaches have been developed to extract insight from the ple
 thora of data available.  As the amount of data grow rapidly, however, cur
 rent approaches for analysis struggle to scale. Thi...\n\n\nThomas Heinis 
 (Imperial College, London)\n---------------------\nAfternoon Keynote - Rob
 inson Pino (DOE ASCR)\n\nRobinson Pino (US Department of Energy Office of 
 Advanced Scientific Computing Research)\n---------------------\nAluminum: 
 An Asynchronous, GPU-Aware Communication Library Optimized for Large-Scale
  Training of Deep Neural Networks on HPC Systems\n\nWe identify communicat
 ion as a major bottleneck for training deep neural networks on large-scale
  GPU clusters, taking over 10x as long as computation. To reduce this over
 head, we discuss techniques to overlap communication and computation as mu
 ch as possible. This leads to much of the communication ...\n\n\nNikoli Dr
 yden (University of Illinois, Lawrence Livermore National Laboratory); Nao
 ya Maruyama, Tim Moon, Tom Benson, and Andy Yoo (Lawrence Livermore Nation
 al Laboratory); Marc Snir (University of Illinois); and Brian Van Essen (L
 awrence Livermore National Laboratory)\n---------------------\nOn Adam-Tra
 ined Models and a Parallel Method to Improve the Generalization Performanc
 e\n\nAdam is a popular stochastic optimizer that uses adaptive estimates o
 f lower-order moments to update weights and requires little hyper-paramete
 r tuning. Some recent studies have called the generalization and out-of-sa
 mple behavior of such adaptive gradient methods into question, and argued 
 that such...\n\n\nGuojing Cong (IBM) and Luca Buratti (IBM, University of 
 Bologna)\n\nTag: Deep Learning, Machine Learning\n\nRegistration Category:
  Workshop Reg Pass
END:VEVENT
END:VCALENDAR