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:20211207T055409Z
LOCATION:222
DTSTART;TZID=America/Chicago:20211115T120000
DTEND;TZID=America/Chicago:20211115T123000
UID:submissions.supercomputing.org_SC21_sess346_ws_indis108@linklings.com
SUMMARY:Learning Transfers via Transfer Learning
DESCRIPTION:Workshop\n\nLearning Transfers via Transfer Learning\n\nArifuz
 zaman, Arslan\n\nDetecting performance anomalies is key to efficiently uti
 lize network resources and improve the quality of service. Researchers pro
 posed various approaches to identify the presence of anomalies by analyzin
 g performance statistics using heuristic (e.g., change point detection) an
 d Machine Learning (ML) models. Although these models yield high accuracy 
 in the networks that they are trained for, their performance degrade sever
 ely when transferred to different network settings. This is because of the
  fact that existing models detect anomalies by capturing the changes in tr
 ansfer throughput and observed RTT values, which are dependent to network 
 settings. In this paper, we propose a novel feature transformation method 
 to eliminate network dependence of ML models for anomaly diagnosis problem
 s to enhance their performance when transferred to new networks, thereby m
 itigating the need to gather training data in each network separately. We 
 validate the findings through experimental evaluations conducted on simula
 ted and production networks and show that the proposed feature transformat
 ion improves the performance of transfer learning for anomaly diagnosis pr
 oblems from less than 60% to over 90%. Finally, we evaluate the performanc
 e of the proposed solutions using various congestion control algorithm and
  observe that the models trained using BBR attains the best transfer learn
 ing performance.\n\nTag: Big Data, Cloud and Distributed Computing, Data A
 nalytics, Data Management, Datacenter, Networks, Performance, Quantum Comp
 uting\n\nRegistration Category: Workshop Reg Pass
END:VEVENT
END:VCALENDAR