Anisotropy

Higher is Better

[0, 1]

Description

Multi-method anisotropy combining log-ellipticity, condition numbers, ratio variance, entropy, dominance, and effective dimensionality

Weighted combination: 0.25*ellipticity + 0.25*condition + 0.20*ratio_var + 0.15*entropy + 0.10*dominance + 0.05*eff_dim

High anisotropy = strong directional bias (good for trajectories). Low anisotropy = spherical distribution (good for steady-state).

Other metrics in the Intrinsic Latent Space (LSE) category

Multi-method dimensionality efficiency score combining variance thresholds, Kaiser criterion, elbow detection, and spectral decay

Rate of eigenvalue decay indicating information concentration in leading dimensions

Effective dimensionality measure from eigenvalue distribution (trajectory: lower is better, steady-state: higher is better)