Neural Gas & SOM — Competitive Learning & Topology

🧠 Neural Gas & SOM — Competitive Learning & Topology AI / ML

Algorithm

Parameters

Learning Rate η 0.30

Neighbourhood σ 2.0

Grid Size 8×8

λ (Neural Gas) 3.0

Data Distribution

Control

Statistics

Epoch0

Quant. Error—

AlgorithmSOM

Nodes—

SOM: Δwᵢ = η·h(d)·(x−wᵢ) where h=e^(−d²/2σ²).
Neural Gas: Δwᵢ = η·e^(−rank(i)/λ)·(x−wᵢ).
σ and η decay over time.

Data distribution

SOM grid (adapting to data)