Just a Few Seeds More: The Inflated Value of Network Data for Diffusion...

on YouTube ↗01h 04m

" summary

Shows that under the classic SIR/independent cascade model, seeding s + x random people often outperforms optimally chosen s seeds: simulations on 75 Indian villages, a Chinese field dataset and Facebook graphs find random seeding with 3 extra seeds beats diffusion-central heuristics and even an omniscient benchmark. Theory proves this for inhomogeneous random graphs when the largest eigenvalue exceeds 1/C (a giant component exists), so the gap vanishes quickly and collecting full network maps is often not cost‑effective.

AI-written summary · about

" chapters (11)

1. 0:00The optimal seeding problem
2. 2:12Applications and heuristics
3. 5:55Empirical motivation
4. 11:26Performance of random seeding
5. 17:04Theoretical framework
6. 20:15Model definition
7. 24:57Main theorem and proof
8. 30:37Percolation perspective
9. 44:11Limitations and extensions
10. 50:04Simulation and RCT evidence
11. 55:03Conclusion and discussion