Module hrp 

Hierarchical Risk Parity (HRP) — pure math functions.

Three-phase algorithm:

1. Compute correlation → distance matrix → agglomerative clustering (average linkage)
2. Build dendrogram (binary tree of clusters)
3. Recursive bisection following dendrogram with inverse-variance weighting

All functions are pure math: &[f64] in, f64 out. No domain types, no I/O. The adapter wraps domain types (e.g. LegReturns) into raw slices before calling these functions.

Enums

ClusterNode

Binary tree node representing hierarchical clustering.

Constants

MIN_OBSERVATIONS

Minimum number of return observations required to compute a meaningful correlation matrix.

Functions

build_dendrogram

Build a dendrogram using agglomerative clustering with weighted average linkage.

cluster_variance

Compute average variance for a cluster (sum of leaf variances / leaf count).

cluster_variance_sum

correlation_matrix

Compute the Pearson correlation matrix from return series.

decimal_from_f64

Convert f64 to Decimal with 6 decimal places precision.

distance_matrix

Convert correlation to distance: d(i,j) = sqrt((1 - corr(i,j)) / 2)

find_closest_pair

mean

recursive_bisect

Walk the dendrogram tree, splitting weight between left and right subtrees using inverse-variance allocation at each branch.

variance