slides - Nicole Immorlica

cs234r
Markets for Networks and Crowds
BRENDAN LUCIER, MICROSOFT RESEARCH NE
NICOLE IMMORLICA, MICROSOFT RESEARCH NE
Lecture 3:
Itai Ashlagi, Yash Kanoria, and Jacob Leshno.
Unbalanced Random Matching Markets: The
Stark Effect of Competition, Journal of Political
Economy, forthcoming.
Unbalanced Markets:
Setting:
𝑛 men, 𝑛 + π‘˜ women, random complete preferences
β€’ Let 𝑅𝑀 πœ‡ = ave. rank of matched men in πœ‡.
β€’ Let π‘…π‘Š πœ‡ = ave. rank of matched women in πœ‡.
π‘š1 : 𝑀1 ≻ 𝑀2 ≻ 𝑀3 ≻ 𝑀4
π‘š2 : 𝑀1 ≻ 𝑀3 ≻ 𝑀4 ≻ 𝑀2
𝑀1 : π‘š3 ≻ π‘š2 ≻ π‘š1
𝑀2 : π‘š1 ≻ π‘š2 ≻ π‘š3
π‘š3 : 𝑀2 ≻ 𝑀4 ≻ 𝑀3 ≻ 𝑀1
𝑀3 : π‘š1 ≻ π‘š3 ≻ π‘š2
𝑀4 : π‘š1 ≻ π‘š2 ≻ π‘š3
π‘˜ = 1,
𝑅𝑀 πœ‡ = 5 3 ,
π‘…π‘Š πœ‡ = 2.
Unbalanced Markets:
Let πœ‡π‘€ (πœ‡π‘Š ) be the man (woman)-opt. matching.
Prior work [Pittel’89], [Knuth-Motwani-Pittel’90]:
Proposing side does well in balanced (π‘˜ = 0) markets.
β€’ Pr 𝑅𝑀 πœ‡π‘€ > 32 ln 𝑛 β†’ 0 as 𝑛 grows
β€’ 𝐸 𝑅𝑀 πœ‡π‘€ ≀ ln 𝑛
β€’ 𝐸 π‘…π‘Š πœ‡π‘€ ≀ 𝑛 ln 𝑛
Unbalanced Markets:
Let πœ‡π‘€ (πœ‡π‘Š ) be the man (woman)-opt. matching.
This paper:
Short side does well in all stable matchings (π‘˜ > 0).
β€’ 𝑅𝑀 πœ‡ ≀ 3 ln π‘›π‘˜ , π‘…π‘Š πœ‡ β‰₯ 3 ln 𝑛𝑛 π‘˜ for any πœ‡
And close to optimal.
β€’ 𝑅𝑀 πœ‡π‘Š ≀ 1 + π‘œ 1 𝑅𝑀 πœ‡π‘€ ,
β€’ π‘…π‘Š πœ‡π‘Š β‰₯ 1 βˆ’ π‘œ 1 π‘…π‘Š πœ‡π‘€
Simulation w/n=40
(the authors vary the # of men; this talk varies # of women)
Simulation w/n=40
(the authors vary the # of men; this talk varies # of women)
Core Idea:
Rejection chains:
Start from man-optimal πœ‡π‘€ .
Iteratively initiate rejection chains to reach πœ‡π‘Š .
Count total # proposals in process.
Analysis:
Let 𝑆 = 𝑀 ∢ πœ‡ 𝑀 best stable mate .
When 𝑀 ∈ 𝑆 receives a proposal, algorithm terminates.
Improvement Cycle
𝑆
𝑆
Analysis:
Let 𝑆 = 𝑀 ∢ πœ‡ 𝑀 best stable mate .
When 𝑀 ∈ 𝑆 receives a proposal, algorithm terminates.
𝑆
Terminal Phase
𝑆
𝑆
Analysis:
Let 𝑆 = 𝑀 ∢ πœ‡ 𝑀 best stable mate .
When a 𝑀 ∈ 𝑆 receives a proposal, can’t be stable.
𝑆
Terminal Phase
𝑆
𝑆
𝑆
𝑆
Analysis:
Consider starting rejection chain from woman 𝑀
who received most proposals in πœ‡π‘€ .
β€’ Phase likely to be terminal since single woman
accepts all proposals, but 𝑀 only accepts
those she prefers to the ones she’s received.
β€’ Pr π‘š picks 𝑀 β‰ˆ Pr[π‘š picks single] = 1 𝑛
β€’ Expected length of chain β‰ˆ 𝑛.
𝑆 large implies future rejection chains are short.
INFORMAL Discussion:
Simulations indicate results hold with correlation.
INFORMAL Discussion:
But advantage diminishes as correlation increases.
INFORMAL Discussion:
Similarly, paper shows impact of
β€’ Many-to-one
β€’ Short lists
Question: What makes a side short?