Class lucid::FourierBarrierCertificate: Add more details to the formulation, as there are some undefined symbols used.
Class lucid::GradientOptimizable: Instead of being a subclass of Estimator, we should introduce a mixin subclass that supports oll kinds of mixins
Class lucid::MultiSet: Simplistic implementation of sampling.
Member lucid::MultiSet::lattice (const VectorI &points_per_dim, bool endpoint) const override: Improve the naive implementation that only concatenates the lattices from the internal sets (polytopes?)
Member lucid::pdist (const Eigen::MatrixBase< Derived > &x): Remove TriangularMatrix tparam in favor of squareform function
Member lucid::Tuner::tune_impl (Estimator &estimator, ConstMatrixRef training_inputs, const OutputComputer &training_outputs) const =0: If the OutputComputer becomes a bottleneck, consider having a separate method for when the outputs are available from the get-go.
Member LUCID_FLAG_ENUMS (name, names, max_enum): For clarity, once the design is finalised, consider moving all the definitions in the correct enum class files.