Learning vector quantization: Difference between revisions

Revision as of 21:40, 18 May 2017

Revision as of 11:04, 23 May 2017

Line 22:

* the corresponding label <math>c_i</math> to each neuron <math> \vec{w_i} </math>

* how fast the neurons are learning <math> \eta </math>

* and an input list containing<math> tuplesL of</math> vectorscontaining withall labelsthe tovectors of trainwich the neuronslabels <math>are Lknown </math>already (Trainingsset).

The algorithm's flow is:

# For next input <math>(\vec{x},c)</math> in <math> L </math> find the neuronsclosest neuron <math>\vec{w_m}</math>, with the same Label and<br>i.e. take the one so that <math>d(\vec{x},\vec{w_m})</math> gets= its\min\limits_i minimum{d(\vec{x},\vec{w_m})} value</math>, where <math>\, d\, </math> is the metric used ( [[Euclidean distance|Euclidean]], etc. ).

# Update <math>\vec{w_m}</math>. A better explanation is get <math>\vec{w_m}</math> closer to the input <math>\vec{x}</math>, if <math>\vec{x}</math> and <math>\vec{w_m}</math> belong to the same label and get them further apart if they don't. <br><math> \vec{w_m} \gets \vec{w_m} + \eta \cdot \left( \vec{x} - \vec{w_m} \right) </math> (closer together) <br> or <math> \vec{w_m} \gets \vec{w_m} - \eta \cdot \left( \vec{x} - \vec{w_m} \right) </math> (further apart).

# While there are vectors left in <math> L </math> go to step 1, else terminate.