Definition. Amplification [amplification]

Modifying the definition of RP\text{RP} a little bit: L∈RP(p)L \in \text{RP}(p) if there is a PTM that, on input xx, runs in time poly(|x|)\text{poly}\left( |x| \right) and has behavior:

  • Pr(reaching accept)≥p\text{Pr}\left( \text{reaching accept} \right) \geq p
  • Pr(reaching reject)=1\text{Pr}\left( \text{reaching reject} \right) = 1

Example. Claim: RP(12)=RP(34)\text{RP}\left( \frac{1}{2} \right) = \text{RP}\left( \frac{3}{4} \right)

Proof.

Let L∈RP(12)L \in \text{RP}\left( \frac{1}{2} \right). There exists PTM MM such that x∈L:P(Macceptx)≥12x \in L:P\left( {M\ \text{accept}\ x} \right) \geq \frac{1}{2}, x∉L:Pr(Mrejectsx)=1x \notin L:\text{Pr}\left( {M\ \text{rejects}\ x} \right) = 1.

Let NN be the Turing machine on input xx

  • Run MM on input xx
  • Run MM on input xx
  • if either accepted, then accept
  • else reject