Locally linear embedding

 

LLE is an unsupervised learning algorithm that computes low-dimensional, neighborhood-preserving embeddings of high-dimensional inputs.

Steps

  1. select neighbors

    \[\boldsymbol{X}_m = \sum_{i\in S_m} W_{mi} \boldsymbol{X}_i\]
  2. reconstruct with weights

    \[\mathcal{F} = \sum^N_{m=1} \| \boldsymbol{X}_m-\sum^N_{i=1} W_{mi} \boldsymbol{X}_i \|^2\]
  3. map to embedded coordinates

    \[\mathcal{G}(\boldsymbol{\xi}_1,...,\boldsymbol{\xi}_N) = \sum^N_{m=1} \|\boldsymbol{\xi}_i-\sum_{i=1}^N W_{mi}\boldsymbol{\xi}_i\|^2\]