Are balls picked uniformly at random? Goldreich

On the Power of Conditional Samples in Distribution Testing
Sourav Chakraborty, Eldar Fischer, Yonatan Goldhirsh and Arie Matsliah
Is the Lottery Fair?
Goldreich-Ron Test:
1. Run 𝑂( 𝑛)
draws
2. Reject if many
collisions found
Collision
Conditional Samples Test:
1. Pick a set H of 𝑂(1) balls
using the machine
2. Pick a set L of 𝑂(1) balls
uniformly at random
3. Test the machine with the
balls in 𝐿 βˆͺ 𝐻
Are balls picked uniformly at random?
Ratio Trees
Explicit Samplers
Summary of Results
Given a conditional samples oracle for a distribution 𝐷
1
3
3
5
2
5
Inner nodes:
2
3
1
2
Adaptive
Produce an oracles for 𝐷 close to 𝐷
which outputs samples and their probability
1
2
Pr 𝑙𝑒𝑓𝑑
Pr π‘ π‘’π‘π‘‘π‘Ÿπ‘’π‘’
Leaves: universe elements
2
5
1
3
2
15
Uniformity
Identity
Label invariant
properties
General
properties
Non-adaptive
1
π‘π‘œπ‘™π‘¦( )
πœ–
1
βˆ—
π‘π‘œπ‘™π‘¦(log 𝑛 , )
πœ–
1
π‘π‘œπ‘™π‘¦(log 𝑛 , )
πœ–
Ξ©(√log log 𝑛)
Ξ© 𝑛
1
π‘π‘œπ‘™π‘¦(log 𝑛 , )
πœ–
Ξ©(log log 𝑛)
Ξ©(log log 𝑛)
(following
uniformity)