Expand description
Data model for the blockwise carrier (subset moved down to gam-problem):
parameter-block specs, the effective-Jacobian and channel-Hessian
abstractions, per-block working sets and states, and the block geometry
directional derivative.
The coefficient-group/label/prior types, custom_family_block_role, and the
blockspec validators remain in the family crate because they depend on
CoefficientGroupPrior/RhoPrior/BlockRole/CustomFamilyError.
Structs§
- Additive
Block Jacobian - A
BlockEffectiveJacobianfor any block that contributes linearly to exactly one output of a multi-output family. - Block
Geometry Directional Derivative - Family
Linearization State - β-linearization state passed to
BlockEffectiveJacobian::effective_jacobian_at. - Gauge
Composed Jacobian - A
BlockEffectiveJacobianthat composes an inner callback’s raw-width effective Jacobian with a fixed reduced→raw block transformT_b(p_raw × r_reduced), so the family sees the reduced coordinates by construction (#933). - Parameter
Block Spec - Static specification for one parameter block in a custom family.
- Parameter
Block State - Current state for a parameter block.
- RowScaled
Jacobian - A
BlockEffectiveJacobianfor a single-output block whose contribution to the linear predictor isdiag(eta_scaling) · design(row-wise scaling). - Tensor
Channel Hessian - A
FamilyChannelHessianbacked directly by a pre-computed(n × K × K)tensor. Used by the defaultchannel_hessian_atimplementation and by tests.
Enums§
- Block
Working Set - Working quantities supplied by a custom family for one block.
Traits§
- Block
Effective Jacobian - β-dependent Jacobian callback for a parameter block.
- Family
Channel Hessian - Per-subject channel Hessian provider for multi-output families.