Function design 

pub fn design<R>(
    hypotheses: usize,
    experiments: &[Experiment<R>],
    objective: MinCostCover<R>,
) -> Result<DesignPlan<R>, QuantumError>
where
    R: RealField + FromPrimitive + Default + Debug,

The exact minimum-cost cover of the C(n, 2) hypothesis pairs by the offered experiments.

A dynamic program over subsets of covered pairs: dp[S | cover(e)] = min(dp[S], dp[S] + cost(e)), relaxed for every state in ascending order and every experiment in declared order. O(2^C(n,2) · k), linear in the experiments and exponential in the hypotheses; enumerating experiment subsets at 2^k is the wrong enumeration and is not what runs. Strict relaxation in declared order breaks every tie the same way, so one instance yields one plan.

An experiment covers a pair when the two hypotheses’ predicted read-outs separate by at least floor_bits at the experiment’s shots, measured as the shot-scaled Bhattacharyya distance and compared with the state member of the tolerance family as slack. Pairs no experiment covers are reported rather than failed: the solve targets the coverable pairs and lists the rest.

Errors

QuantumError::HypothesisCountExceeded naming n and C(n, 2) when n exceeds objective.max_hypotheses or C(n, 2) exceeds the pair mask, before the pairs or the table are allocated; C(n, 2) is computed in checked arithmetic, and an overflow is reported as usize::MAX pairs. QuantumError::CalculationError if fewer than two hypotheses are offered, since there is then no pair to cover, or if objective.floor_bits is not finite or is negative. QuantumError::DimensionMismatch if an experiment predicts for a different number of hypotheses.