liveBook · Manning

Gaussian neighborhood function

As explained in Build AI Drug Discovery Pipelines

A Gaussian neighborhood function assigns larger update weights to nodes nearer the best matching unit and smaller weights to nodes farther away according to a Gaussian profile.

A Gaussian neighborhood function weights self-organizing map updates by distance from the best matching unit (BMU): nearby nodes move more, while distant nodes move less. Its width, σ(t), and the learning rate typically decrease over time, shifting learning from broad organization toward fine-tuning. section 7.1.2

Use the function to:

  • Adjust each node toward an input vector according to its distance from the BMU. section 7.1.2
  • Organize similar high-dimensional compound fingerprints into nearby grid regions. section 7.1.2

Compound mapping with MiniSom

A MiniSom configuration selects the Gaussian neighborhood function for a rectangular grid and evaluates candidate σ, learning-rate, and activation-distance settings on fingerprint data.

How Build AI Drug Discovery Pipelines covers it

How deep it goes
Explains distance-based weighting, parameter decay, and grid interpretation, but does not derive the function beyond its equation.
Examples
Uses MiniSom parameter search, Morgan-fingerprint compounds, rectangular grids, and node-level cluster visualizations.
What it assumes
Assumes vectorized inputs and BMU selection while defining how neighborhood distance controls node updates.
What you can do after
Apply it via MiniSom by configuring Gaussian neighborhoods and tuning width, learning rate, and activation distance.
How it connects
Connects Gaussian neighborhoods to self-organizing maps, best matching units, learning rates, and high-dimensional molecular fingerprints.

Related concepts

This explainer was generated by Manning's knowledge graph from the text of Build AI Drug Discovery Pipelines. The section links point to the parts of the book it draws on.