Finite seeded extraction and sampling from weak sources #
The sampler tests only small output sets, as in Chattopadhyay and Liao, Extractors for Sum of Two Sources (2021), Definition 3.14. Its source may have any probability weights with the stated point-mass bound. The seeded extractor premise only needs uniform sources on finite supports, since the proof applies it to the full set of bad inputs.
Seed and output types are assumed nonempty by the public conversion theorems. All counts retain seed multiplicity, including repeated output values.
Number of seeds whose output belongs to the test set, at a fixed input.
Equations
- Algebraic.Cutwidth.Extractor.seedHits E T x = {y : Seed | E x y ∈ T}.card
Instances For
Inputs for which more than a 2ε fraction of seeds hit the test set.
Equations
- Algebraic.Cutwidth.Extractor.badInputs E T ε = {x : α | 2 * ε * ↑(Fintype.card Seed) < ↑(Algebraic.Cutwidth.Extractor.seedHits E T x)}
Instances For
Seeded extraction from every uniform support of size at least K.
The discrepancy for every output test is multiplied by the number of source
pairs and output values, so this definition uses no division.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Small-set sampling for every probability source with point masses at
most 1 / L, expressed as L * p x ≤ 1. For positive L, this is source
min-entropy at least log₂ L. The seed is uniform and independent of the source.
Equations
- One or more equations did not get rendered due to their size.