BEGIN:VCALENDAR
VERSION:2.0
PRODID:IEEE vTools.Events//EN
CALSCALE:GREGORIAN
BEGIN:VTIMEZONE
TZID:Asia/Shanghai
BEGIN:STANDARD
DTSTART:19910915T010000
TZOFFSETFROM:+0900
TZOFFSETTO:+0800
TZNAME:CST
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20261014T025859Z
UID:A729836F-F168-48D5-9CEA-172387C6A568
DTSTART;TZID=Asia/Shanghai:20261017T100000
DTEND;TZID=Asia/Shanghai:20261017T110000
DESCRIPTION:Traditional optimization algorithms rely on repeated manual red
 esign. This talk proposes a new paradigm of Algorithm Modelization\, which
  treats algorithm execution trajectories as computational experience and b
 uilds the A2M model. A two‑stage “imitation‑to‑surpassing” train
 ing integrates knowledge from evolutionary\, Bayesian and gradient‑based
  optimizers. The model shows strong generalization across problems and ext
 rapolates to higher unseen dimensions. It can keep absorbing new algorithm
 s for progressive enhancement. Experiments validate this paradigm\, paving
  the way toward general optimization models that accumulate and evolve wit
 hout hand‑crafted rules.\n\nSpeaker(s): Kay Chen Tan\, \n\nGuangzhou\, G
 uangdong\, China
LOCATION:Guangzhou\, Guangdong\, China
ORGANIZER:csziboqin@mail.scut.edu.cn
SEQUENCE:0
SUMMARY:Algorithm Modelization: Evolving Optimization Algorithms into Learn
 able and Evolvable Intelligence
URL;VALUE=URI:https://events.vtools.ieee.org/m/585084
X-ALT-DESC:Description: <br /><p>Traditional optimization algorithms rely o
 n repeated manual redesign. This talk proposes a new paradigm of Algorithm
  Modelization\, which treats algorithm execution trajectories as computati
 onal experience and builds the A2M model. A two‑stage &ldquo\;imitation
 ‑to‑surpassing&rdquo\; training integrates knowledge from evolutionary
 \, Bayesian and gradient‑based optimizers. The model shows strong genera
 lization across problems and extrapolates to higher unseen dimensions. It 
 can keep absorbing new algorithms for progressive enhancement. Experiments
  validate this paradigm\, paving the way toward general optimization model
 s that accumulate and evolve without hand‑crafted rules.</p>
END:VEVENT
END:VCALENDAR