Fully tolerant product state testing and closest product state learning

Abstract

We address the problem of testing whether an unknown $n$-qudit state $ρ$ is $a$-close to a product state or $b$-far away from any product state, as measured in terms of state overlap. We provide a time-efficient algorithm to solve this problem that requires an $n$-independent number of copies of the unknown state. Our random coloring argument shows that for any unknown state there always exists a partition of $[n]$ into $q$ parts such that the square of the overlap with the closest product state with respect to this partition is only an additive factor $O(1/q)$ larger. This reduces the problem to tolerant testing of $q$ parties, with potentially growing local dimensions. Combining this insight with blockwise spectral projection arguments we can show that the natural $k$-copy generalization of Harrow & Montanaro's product state test provides an efficient tolerant tester. We use the same random coloring and blockwise spectral projection techniques to obtain a substantially improved algorithm for closest product state learning. In particular we give an algorithm that takes in $\widetilde{O}\big((nd)^2\big)\, 2^{\widetilde{O}(1/\varepsilon^8)}$ copies of the unknown state and produces an $\varepsilon$-approximately optimal product state. The key technical components of this learner are a qudit variant of the Bakshi et al. high-fidelity product state learning algorithm and a sampling technique based on Werner's optimal cloning channel.