Solo reading/watching: Fefferman’s talk at QIP (PreciseQMA = PSPACE)

Consider the precise succinct hamiltonian problem:

• “succinct encoding” = can get elements of row of $$2^k \times 2^k$$ matrix in $$poly(n)$$ time, $$k(n)$$ space
• generalization of sparsity condition (when $$k(n)$$ is poly(n))
• “precise” -> inverse exponential gap $$2^{-O(k(n))}$$ (usually inverse polynomial)

This is complete for PreciseQMA! (precise + quantum NP) = BQPSPACE

This is the power of “small gaps” + “quantumness” – both are needed to generalize to PreciseQMA.

Consider BQPSPACE = PSPACE

• BQPSPACE$$(k)$$ is bounded error quantum algorithm, $$k(n)$$ qubits space, at most exponential runtime
• Promise problem, but error can be inverse polynomial with amplitude amplification

Another result: consider unitary quantum space (non-intermediate measurements only!)

• recent paper says this is the same as non-unitary BQPSPACE, you don’t need to add ancillas!

• k-local hamiltonian problem is qma complete (verify yes/no with quantum circuit, above some error)

• can do well conditioned matrix inversion.

##### Open questions:
• What’s QMA vs QMA(2)? Where you have poly vs two Merlins that give you a result to verify?
• How does QMA(k) differ from QIP(k)?