BEGIN:VCALENDAR VERSION:2.0 PRODID:-//Drupal iCal API//EN X-WR-CALNAME:Events items teaser BEGIN:VEVENT UID:62d3df227c7b0 DTSTART;TZID=America/Toronto:20181214T150000 SEQUENCE:0 TRANSP:TRANSPARENT DTEND;TZID=America/Toronto:20181214T150000 SUMMARY:PhD Seminar: Progressive Memory Banks for Incremental Domain\nAdapt ation CLASS:PUBLIC DESCRIPTION:Summary \n\nNABIHA ASGHAR\, PHD CANDIDATE\n_David R. Cheriton S chool of Computer Science_\n\nWe address the problem of incremental domain adaptation (IDA). We\nassume each domain comes one after another\, and th at we could only\naccess data in the current domain. The goal of IDA is t o build a\nunified model performing well on all the domains that we have\n encountered. We propose to augment a recurrent neural network (RNN)\nwith a directly parameterized memory bank\, which is retrieved by an\nattention mechanism at each step of RNN transition. The memory bank\nprovides a nat ural way of IDA: when adapting our model to a new\ndomain\, we progressive ly add new slots to the memory bank\, which\nincreases the number of param eters\, and thus the model capacity. \n DTSTAMP:20220717T100626Z END:VEVENT END:VCALENDAR