lle-single-hyperparameter-no-internal-model

IN premise — summaries/2026/08/24/wiki-Nonlinear_dimensionality_reduction.md

Created 2026-08-24T17:11:20+00:00

Locally-Linear Embedding (LLE) has exactly one hyperparameter (K, the number of nearest neighbors) and has no internal model, meaning it cannot project new out-of-sample points into an existing embedding without additional regularization

Summary

LLE is a one-shot, batch-only method with a single tunable knob (how many neighbors to use), so it produces a fixed embedding and offers no mechanism for mapping new data points into that same space. In practice, this means any system built around LLE must either recompute the entire embedding whenever new observations arrive or bolt on a separate regularization step, making it ill-suited for streaming or incremental pipelines.