BEGIN:VCALENDAR
VERSION:2.0
PRODID:icalendar-ruby
CALSCALE:GREGORIAN
METHOD:PUBLISH
BEGIN:VTIMEZONE
TZID:Europe/Vienna
BEGIN:DAYLIGHT
DTSTART:20260329T030000
TZOFFSETFROM:+0100
TZOFFSETTO:+0200
RRULE:FREQ=YEARLY;BYDAY=-1SU;BYMONTH=3
TZNAME:CEST
END:DAYLIGHT
BEGIN:STANDARD
DTSTART:20251026T020000
TZOFFSETFROM:+0200
TZOFFSETTO:+0100
RRULE:FREQ=YEARLY;BYDAY=-1SU;BYMONTH=10
TZNAME:CET
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20260308T203731Z
UID:699c1168a3e99368223411@ist.ac.at
DTSTART:20260305T113000
DTEND:20260305T123000
DESCRIPTION:Speaker: Joel Daniel Andersson\nAbstract: We study differential
 ly-private statistics in the fully dynamic continual observation model\, w
 here many updates can arrive at each time step\, and updates to a stream c
 an involve both insertions and deletions of an item. Earlier work (e.g.\, 
 Jain et al.\, NeurIPS 2023 for counting distinct elements\; Raskhodnikova 
 & Steiner\, PODS 2025 for triangle counting with edge updates) reduced the
  respective cardinality estimation problem to continual counting on the di
 fference stream associated with the true function values on the input stre
 am.In such reductions\, a change in the original stream can cause many cha
 nges in the difference stream.This poses a challenge for applying private 
 continual counting algorithms to obtain optimal error bounds.Our work impr
 oves the accuracy of several such reductions by studying the induced neigh
 boring relation more closely.The improvement stems from tight analysis of 
 known factorization mechanisms for the prefix-sum matrix in this setting. 
 In particular\, we show how the square-root factorization (Henzinger et al
 .\, SODA 2023\; Fichtenberger et al.\, ICML 2023) can be employed to give 
 concrete improvements in accuracy over past work based on the binary tree 
 mechanism.We instantiate our framework for the problems of counting distin
 ct elements\, estimating degree histograms\, and estimating triangle count
 s (under a slightly relaxed privacy model)\, showing improved accuracy for
  a large range of parameters.Based on joint work with Palak Jain and Satch
 it Sivakumar (https://arxiv.org/abs/2601.02257).
LOCATION:Moonstone Bldg / Ground floor / Seminar Room G (I24.EG.030g)\, IST
 A
ORGANIZER:achaturv@ist.ac.at
SUMMARY:Joel Daniel Andersson: Improved Accuracy for Private Continual Card
 inality Estimation in Fully Dynamic Streams via Matrix Factorization
URL:https://talks-calendar.ista.ac.at/events/6318
END:VEVENT
END:VCALENDAR