LVQ1 Classifier

Implementation of a learning vector quantization classifier in python 2.7.

The LVQ classifier can be used by importing the class Lvq1 from the lvq1.py file.

Note that only the standard LVQ classifier also known as LVQ1 is implemented. Other versions of LVQ such as OLVQ1 or LVQ3 will be added in potential future updates.

Required libraries

• SciPy
• NumPy

How to use

This is an example of how to use the LVQ model in code.

from lvq.lvq1 import Lvq1
from lvq.lvq_exceptions import InvalidParameterException

training_data , target_labels = getTrainingData() # Get list of training samples and label list
validation_data, validation_labels = getValidationData() # Get list of validation samples and label list


neuron_counts = [5, 5] # Number of neurons per class
labels = [0, 1] # Label for each class
max_epochs = 100
patience = 5 # Number of epochs to wait before stopping training
init_lr = 0.1 # Initial learning rate

try:
  # Create model object
  lvq_network = Lvq1([5,5], [0, 1], max_epochs, lr_hyperparam, patience)
  lvq_network.set_early_stopping(True)
  
  # Manually set validation instead of randomly splitting the training set
  lvq_network.set_validation_set(validation_data, validation_labels)
  
  # Train by initializing neuron weights with values of samples in the training data
  lvq_network.train(training_data, target_labels, "real-samples")  
  
except InvalidParameterException as exception:
  print(exception)