1 of 14

Approximating One-Sided and Two-Sided Nash Social Welfare With Capacities

Salil Gokhale

IIT Delhi

Rohit Vaish

IIT Delhi

Harshul Sagar

IIT Delhi

Jatin Yadav

IIT Delhi

Joint work with:

2 of 14

One-sided Model

 

 

 

 

Valuation function

3 of 14

One-sided Model with Capacities

 

 

 

 

4 of 14

Two-sided Model

Valuation function

Valuation function

 

 

5 of 14

Two-sided Model with Capacities

 

 

 

 

6 of 14

Classes of Valuation Functions

  • Additive:

  • Submodular:

  • Subadditive:

  • Monotone:

7 of 14

Previous Results

1. Jugal Garg, Edin Husi´c, Wenzheng Li, László A Végh, and Jan Vondrák. Approximating Nash Social Welfare by Matching and Local Search. STOC, 2023

2. Pallavi Jain and Rohit Vaish. Maximizing Nash Social Welfare under Two-Sided Preferences. AAAI, 2024

Hardness

Algorithm

e/(e-1)

One-sided Model

Without

Capacities

[Garg et al.,2023]

With

Capacities

Hardness

Algorithm

Two-sided Model

Without

Capacities

[Jain and Vaish, 2024]

With

Capacities

4

[Garg et al.,2023]

Submodular

Additive

8 of 14

Our Results

1. Jugal Garg, Edin Husi´c, Wenzheng Li, László A Végh, and Jan Vondrák. Approximating Nash Social Welfare by Matching and Local Search. STOC, 2023

Hardness

Algorithm

6

One-sided Model

Without

Capacities

With

Capacities

Hardness

Algorithm

Two-sided Model

Without

Capacities

With

Capacities

1.0000759

(even for additive)

1.33

e/(e-1)

[Garg et al.,2023]

4

[Garg et al.,2023]

Submodular

Subadditive

9 of 14

4-Approximate Algorithm for One-sided NSW without Capacities

[Garg et al.,2023]

 

6-Approximate Algorithm for One-sided NSW with Capacities

Cardinality preserving!

Assign the remaining items using

a local search with two-way swaps

10 of 14

Algorithm for Two-sided NSW without Capacities

Phase 1: Find a Nash optimal one to one matching.

Phase 2: Assign each remaining worker to their favourite firm.

A 1.33-approximation!

11 of 14

Algorithm for Two-sided NSW with Capacities

MIN-COST-FLOW does Phase 1 and Phase 2 together.

12 of 14

Open Questions

  • Close the constant factor gaps for Nash Social Welfare.

  • Approximate two-sided weighted Nash Social Welfare

13 of 14

Open Questions

  • Close the algorithm and hardness gaps for Nash Social Welfare.

  • Approximate two-sided weighted Nash Social Welfare

  • Economic properties of NSW in the two sided model?

14 of 14

Thank You! Questions?