BEGIN:VCALENDAR
VERSION:2.0
PRODID:icalendar-ruby
CALSCALE:GREGORIAN
X-WR-CALNAME:Final Doctoral Defense for S M Shovan
X-WR-TIMEZONE:Central Time (US & Canada)
BEGIN:VEVENT
DTSTAMP:20260716T225850Z
UID:tag:localist.com\,2008:EventInstance_53311043233395
DTSTART:20260710T180000Z
DTEND:20260710T200000Z
DESCRIPTION:S M Shovan\, a doctoral candidate in computer science\, will de
 fend their dissertation titled “Parallel Algorithms and Data Structure f
 or Large-Scale Graph Analytics.” Their advisor\, Dr. Sajal K. Das is an 
 curators distinguished professor in computer science. The dissertation abs
 tract is provided below.\n\nReal-world graphs evolve continuously. Protein
 s bind and unbind in response to cellular signals\, sensors join and leave
  networks as batteries deplete\, roads close due to construction or disast
 er\, and citation graphs grow as papers are published. As the graph change
 s\, any property computed on its earlier state becomes stale. Recomputing 
 these properties from scratch after every change is the obvious baseline\,
  but it is catastrophically inefficient at the scales modern applications 
 demand.\n\nThis dissertation develops scalable parallel algorithms for mai
 ntaining graph and hy-pergraph properties efficiently under continuous top
 ological change. The unifying strategy is a two-step identify-and-update f
 ramework: given a batch of edge insertions and deletions\, the framework i
 dentifies the affected subgraph in parallel\, then iteratively propagates 
 prop-erty updates within that bounded region until convergence. This princ
 iple avoids global recomputation and scales to networks with millions of v
 ertices and billions of edges on shared-memory multi-core CPUs and massive
 ly parallel GPU architectures.\n\nWe apply this framework to four graph pr
 operties across six algorithms. For multi- objective shortest paths (MOSP)
 \, DynaMOSP achieves average speedups of 12.74→ on shared-memory CPUs an
 d 69.19→ on Nvidia GPUs over state-of-the-art baselines\, with peak spee
 dups of 57.22→ and 105.39→\, respectively. For dynamic hypergraph tria
 d count-ing\, ESCHER\, the first GPU-based data structure for large dynami
 c hypergraphs\, achieves speedups of up to 104.5→\, 473.7→\, and 112.5
 → across three triad categories. Its memory- efficient successor ESCHER+
  achieves approximately 110→ speedup with a significantly reduced memory
  footprint. For dynamic label propagation in graph-based semi-supervised l
 earning\, DynLP achieves an average speedup of 13→ (up to 102→) over s
 tate-of-the-art methods and up to 100→ better memory efficiency than den
 se harmonic-solution approaches.
GEO:37.955863;-91.774411
LOCATION:Computer Science Building\, 104
SUMMARY:Final Doctoral Defense for S M Shovan
URL;VALUE=URI:https://calendar.mst.edu/event/final-doctoral-defense-for-s-m
 -shovan
END:VEVENT
END:VCALENDAR