BEGIN:VCALENDAR VERSION:2.0 PRODID:IEEE vTools.Events//EN CALSCALE:GREGORIAN BEGIN:VTIMEZONE TZID:America/Regina BEGIN:STANDARD DTSTART:19600424T030000 TZOFFSETFROM:-0700 TZOFFSETTO:-0600 TZNAME:CST END:STANDARD END:VTIMEZONE BEGIN:VEVENT DTSTAMP:20201117T211408Z UID:2162B94C-07E8-4B2D-AE99-621A63B96A80 DTSTART;TZID=America/Regina:20201117T140000 DTEND;TZID=America/Regina:20201117T150000 DESCRIPTION:Abstract: In systems with multiple potentially deceptive agents \, any single agent may have to assess the trustworthiness of other agents in order to decide with which agents to interact. In this context\, indir ect trust refers to trust established through third-party advice\, while d irect trust is derived from one’s own interactions.\n\nSince the adviser s themselves may be deceptive or unreliable\, agents need a mechanism to a ssess and properly incorporate advice. This presentation introduces a new and easy to implement method for computing indirect trust\, based on a sim ple prediction with expert advice strategy as is often used in sequential learning. The new method either is on par with or outperforms competing st ate-of-the-art systems in the vast majority of settings simulated in an em pirical study\, while scaling substantially better. The results demonstrat e that existing systems for computing indirect trust are overly complex\; the problem can be solved much more efficiently than the literature sugges ts.\n\nA second experimental study shows that strong methods for computing indirect trust make direct trust redundant in a surprisingly wide variety of scenarios. Further\, a new method for the combination of the two trust types is proposed that\, in the remaining scenarios\, outperforms the one s known from the literature.\n\nThis is joint work with Elham Parhizkar\, Mohammad Hossein Nikravan\, and Robert C. Holte.\n\nSpeaker(s): Dr. Sandra Zilles\, \n\nAgenda: \nWhen: Tuesday\, 17 November 2020\n\nTime: 2:00 – 3:00 pm (Regina Time)\n\nWhere: Zoom Presentation (Link provided in locat ion section)\n\nRegina\, Saskatchewan\, Canada\, Virtual: https://events.v tools.ieee.org/m/242465 LOCATION:Regina\, Saskatchewan\, Canada\, Virtual: https://events.vtools.ie ee.org/m/242465 ORGANIZER:moeeed@ieee.org SEQUENCE:3 SUMMARY:Webinar: Efficient Methods for Computing Trust in Multi-Agent Syste ms URL;VALUE=URI:https://events.vtools.ieee.org/m/242465 X-ALT-DESC:Description: <br /><p><strong>Abstract:</strong><strong>&nbsp\; </strong>In systems with multiple potentially deceptive agents\, any singl e agent may have to assess the trustworthiness of other agents in order to decide with which agents to interact. In this context\, indirect trust re fers to trust established through third-party advice\, while direct trust is derived from one&rsquo\;s own interactions.</p>\n<p>Since the advisers themselves may be deceptive or unreliable\, agents need a mechanism to ass ess and properly incorporate advice. This presentation introduces a new an d easy to implement method for computing indirect trust\, based on a simpl e prediction with expert advice strategy as is often used in sequential le arning. The new method either is on par with or outperforms competing stat e-of-the-art systems in the vast majority of settings simulated in an empi rical study\, while scaling substantially better. The results demonstrate that existing systems for computing indirect trust are overly complex\; th e problem can be solved much more efficiently than the literature suggests .</p>\n<p>A second experimental study shows that strong methods for comput ing indirect trust make direct trust redundant in a surprisingly wide vari ety of scenarios. Further\, a new method for the combination of the two tr ust types is proposed that\, in the remaining scenarios\, outperforms the ones known from the literature.</p>\n<p>This is joint work with Elham Parh izkar\, Mohammad Hossein Nikravan\, and Robert C. Holte.</p><br /><br />Ag enda: <br /><p><strong>When</strong>: Tuesday\, 17 November 2020</p>\n<p>< strong>Time</strong>: 2:00 &ndash\; 3:00 pm (Regina Time)</p>\n<p><strong> Where</strong>: Zoom Presentation (Link provided in location section)</p> END:VEVENT END:VCALENDAR