Now 6199 visitors
Today:0 Yesterday:0
Total: 152377 212S 50P 4R
2026-10-20, Week 43
Member Login
TACT Journal Homepage
Call for Paper
Committee
Find My Paper
Author Homepage
Statistics & History
Welcome Message
Paper Procedure
FAQ
Paper Archives
Outstanding Papers
Program & Proceedings
Presentation Platform
Hotel & Travel Info
Photo Gallery
Scheduler Login
Seminar
Archives Login
Sponsors




















IEEE/ICACT20260329 Question.6
Questioner: tanghr@dsp.ac.cn    2026-02-08 ¿ÀÈÄ 2:26:00
IEEE/ICACT20260329 Answer.6
Answer by Auhor yangyn@dsp.ac.cn   2026-02-08 ¿ÀÈÄ 2:26:00
How does the compressed state summary (SS) in CDynP prevent privacy leaks while still enabling optimal global strategy generation? Does it truly not expose any internal information? The compressed state summary (SS) in CDynP prevents privacy leaks by design: it only transmits the minimal maximum stage execution time for each stage-layer combination, computed as SSk−1​[s][l]=ming​dpk−1​[s][l][g], which aggregates cost metrics without exposing internal GPU allocations, topology, or resource details. This compressed data enables global optimization by allowing each data center to perform local dynamic programming based on inherited SS from predecessors, thus coordinating strategy generation along a chain. The papers theoretical analysis confirms that CDynP achieves near-optimal solutions equivalent to full-information methods under privacy constraints, as SS excludes sensitive intra-DC data like bandwidth or GPU mappings. Thus, it truly exposes no internal information while maintaining training efficiency.

Select Voice