class neupy.algorithms.LVQ3[source]

Learning Vector Quantization 3 (LVQ3) algorithm. Improved version for the LVQ2.1 algorithm.

n_inputs : int

Number of input units. It should be equal to the number of features in the input data set.

n_subclasses : int, None

Defines total number of subclasses. Values should be greater or equal to the number of classes. None will set up number of subclasses equal to the number of classes. Defaults to None (or the same as n_classes).

n_classes : int

Number of classes in the data set.

prototypes_per_class : list, None

Defines number of prototypes per each class. For instance, if n_classes=3 and n_subclasses=8 then there are can be 3 subclasses for the first class, 3 for the second one and 2 for the third one (3 + 3 + 2 == 8). The following example can be specified as prototypes_per_class=[3, 3, 2].

There are two rules that apply to this parameter:

  1. sum(prototypes_per_class) == n_subclasses
  2. len(prototypes_per_class) == n_classes

The None value will distribute approximately equal number of subclasses per each class. It’s approximately, because in casses when n_subclasses % n_classes != 0 there is no way to distribute equal number of subclasses per each class.

Defaults to None.

epsilon : float

Ration between to closest subclasses that triggers double weight update. Defaults to 0.1.

slowdown_rate : float

Paremeter scales learning step in order to decrease it in case if the two closest subclasses predict target value correctly. Defaults to 0.4.

step : float

Learning rate, defaults to 0.01.

show_epoch : int or str

This property controls how often the network will display information about training. There are two main syntaxes for this property.

  • You can define it as a positive integer number. It defines how offen would you like to see summary output in terminal. For instance, number 100 mean that network shows summary at 100th, 200th, 300th … epochs.
  • String defines number of times you want to see output in terminal. For instance, value '2 times' mean that the network will show output twice with approximately equal period of epochs and one additional output would be after the finall epoch.

Defaults to 1.

shuffle_data : bool

If it’s True class shuffles all your training data before training your network, defaults to True.

epoch_end_signal : function

Calls this function when train epoch finishes.

train_end_signal : function

Calls this function when train process finishes.

verbose : bool

Property controls verbose output interminal. True enables informative output in the terminal and False - disable it. Defaults to False.


  • Input data needs to be normalized, because LVQ uses Euclidian distance to find clusters.
  • Training error is just a ratio of miscassified samples
  • Decreasing step and increasing number of training epochs can improve the performance.


>>> import numpy as np
>>> from neupy import algorithms
>>> X = np.array([[0, 0], [0, 1], [1, 0], [1, 1], [2, 2], [1, 2]])
>>> y = np.array([0, 0, 0, 1, 1, 1])
>>> lvqnet = algorithms.LVQ3(n_inputs=2, n_classes=2)
>>> lvqnet.train(X, y, epochs=100)
>>> lvqnet.predict([[2, 1], [-1, -1]])
array([1, 0])
options = {'epoch_end_signal': Option(class_name='BaseNetwork', value=Property(name="epoch_end_signal")), 'epsilon': Option(class_name='LVQ2', value=NumberProperty(name="epsilon")), 'minstep': Option(class_name='LVQ', value=NumberProperty(name="minstep")), 'n_classes': Option(class_name='LVQ', value=IntProperty(name="n_classes")), 'n_inputs': Option(class_name='LVQ', value=IntProperty(name="n_inputs")), 'n_subclasses': Option(class_name='LVQ', value=IntProperty(name="n_subclasses")), 'n_updates_to_stepdrop': Option(class_name='LVQ', value=IntProperty(name="n_updates_to_stepdrop")), 'prototypes_per_class': Option(class_name='LVQ', value=TypedListProperty(name="prototypes_per_class")), 'show_epoch': Option(class_name='BaseNetwork', value=ShowEpochProperty(name="show_epoch")), 'shuffle_data': Option(class_name='BaseNetwork', value=Property(name="shuffle_data")), 'slowdown_rate': Option(class_name='LVQ3', value=NumberProperty(name="slowdown_rate")), 'step': Option(class_name='LVQ3', value=NumberProperty(name="step")), 'train_end_signal': Option(class_name='BaseNetwork', value=Property(name="train_end_signal")), 'verbose': Option(class_name='Verbose', value=VerboseProperty(name="verbose")), 'weight': Option(class_name='LVQ', value=Property(name="weight"))}[source]
slowdown_rate = None[source]
step = None[source]
train_epoch(input_train, target_train)[source]