Bounded-Loss Private Prediction Markets Rafael Frongillo - - PowerPoint PPT Presentation

▶

Mar 19, 2023 133 likes •302 views

Bounded-Loss Private Prediction Markets Rafael Frongillo University of Colorado, Boulder Neural Information Bo Waggoner Microsoft Research, NYC Processing Systems, December 2018 1 / 74 Prediction markets prediction p (0) learning

SLIDE 1

Bounded-Loss Private Prediction Markets

Rafael Frongillo University of Colorado, Boulder Bo Waggoner Microsoft Research, NYC Neural Information Processing Systems, December 2018

1 / 74

SLIDE 2

Prediction markets

learning (aggregation) algorithm time

?

future event

prediction p(0)

2 / 74

SLIDE 3

Prediction markets

learning (aggregation) algorithm time

?

future event

Menu of purchases (gambles)

3 / 74

SLIDE 4

Prediction markets

learning (aggregation) algorithm time

?

future event

purchase choice

4 / 74

SLIDE 5

Prediction markets

learning (aggregation) algorithm time

?

future event

updated prediction p(1)

5 / 74

SLIDE 6

Prediction markets

learning (aggregation) algorithm time

?

future event

updated prediction p(2)

6 / 74

SLIDE 7

Prediction markets

learning (aggregation) algorithm time

?

future event

updated prediction p(t)

7 / 74

SLIDE 8

Prediction markets

time

bserve

event

8 / 74

SLIDE 9

Prediction markets

time

bserve

event payments

9 / 74

SLIDE 10

Prior work

Abernethy-F.-W. 2015 (Neural Information Processing Systems)

10 / 74

SLIDE 11

Prior work

Abernethy-F.-W. 2015 (Neural Information Processing Systems)
differentially private prediction markets

11 / 74

SLIDE 12

Prior work

Abernethy-F.-W. 2015 (Neural Information Processing Systems)
differentially private prediction markets
financial loss may not be bounded!

12 / 74

SLIDE 13

Prior work

Abernethy-F.-W. 2015 (Neural Information Processing Systems)
differentially private prediction markets
financial loss may not be bounded!
Cummings, Pennock, Wortman Vaughan 2016 (EC)
Impossibility: all private market makers

have unbounded financial loss!

13 / 74

SLIDE 14

This paper

Main result: a market construction with:

differential privacy
incentives to participate
accuracy/fidelity of predictions
Bounded worst-case financial loss

14 / 74

SLIDE 15

This paper

Main result: a market construction with:

differential privacy (ε)
incentives to participate (α)
accuracy/fidelity of predictions (α , dimension d)
Bounded worst-case financial loss Õ(d / ε α)

Extensions (cf AFW’15): purchasing data, kernel methods….

15 / 74

SLIDE 16

This paper

Main result: a market construction with:

differential privacy (ε)
incentives to participate (α)
accuracy/fidelity of predictions (α , dimension d)
Bounded worst-case financial loss Õ(d / ε α)

Extensions (cf AFW’15): purchasing data, kernel methods…. Thanks!

74 / 74