pyclustering.nnet.som.som Class Reference

Represents self-organized feature map (SOM). More...

Public Member Functions
def	size (self)
	Return size of self-organized map that is defined by total number of neurons. More...
def	weights (self)
	Return weight of each neuron. More...
def	awards (self)
	Return amount of captured objects by each neuron after training. More...
def	capture_objects (self)
	Returns indexes of captured objects by each neuron. More...
def	__init__ (self, rows, cols, conn_type=type_conn.grid_eight, parameters=None, ccore=True)
	Constructor of self-organized map. More...
def	__del__ (self)
	Destructor of the self-organized feature map.
def	__len__ (self)
	Returns size of the network that defines by amount of neuron in it. More...
def	__getstate__ (self)
def	__setstate__ (self, som_state)
def	train (self, data, epochs, autostop=False)
	Trains self-organized feature map (SOM). More...
def	simulate (self, input_pattern)
	Processes input pattern (no learining) and returns index of neuron-winner. More...
def	get_winner_number (self)
	Calculates number of winner at the last step of learning process. More...
def	show_distance_matrix (self)
	Shows gray visualization of U-matrix (distance matrix). More...
def	get_distance_matrix (self)
	Calculates distance matrix (U-matrix). More...
def	show_density_matrix (self, surface_divider=20.0)
	Show density matrix (P-matrix) using kernel density estimation. More...
def	get_density_matrix (self, surface_divider=20.0)
	Calculates density matrix (P-Matrix). More...
def	show_winner_matrix (self)
	Show a winner matrix where each element corresponds to neuron and value represents amount of won objects from input data-space at the last training iteration. More...
def	show_network (self, awards=False, belongs=False, coupling=True, dataset=True, marker_type='o')
	Shows neurons in the dimension of data. More...

Detailed Description

Represents self-organized feature map (SOM).

The self-organizing feature map (SOM) method is a powerful tool for the visualization of of high-dimensional data. It converts complex, nonlinear statistical relationships between high-dimensional data into simple geometric relationships on a low-dimensional display.

ccore option can be specified in order to control using C++ implementation of pyclustering library. By default C++ implementation is on. C++ implementation improves performance of the self-organized feature map.

Example:

import random

from pyclustering.utils import read_sample
from pyclustering.nnet.som import som, type_conn, type_init, som_parameters
from pyclustering.samples.definitions import FCPS_SAMPLES

# read sample 'Lsun' from file
sample = read_sample(FCPS_SAMPLES.SAMPLE_LSUN)

# create SOM parameters
parameters = som_parameters()

# create self-organized feature map with size 7x7
rows = 10  # five rows
cols = 10  # five columns
structure = type_conn.grid_four;  # each neuron has max. four neighbors.
network = som(rows, cols, structure, parameters)

# train network on 'Lsun' sample during 100 epouchs.
network.train(sample, 100)

# simulate trained network using randomly modified point from input dataset.
index_point = random.randint(0, len(sample) - 1)
point = sample[index_point]  # obtain randomly point from data
point[0] += random.random() * 0.2  # change randomly X-coordinate
point[1] += random.random() * 0.2  # change randomly Y-coordinate
index_winner = network.simulate(point)

# check what are objects from input data are much close to randomly modified.
index_similar_objects = network.capture_objects[index_winner]

# neuron contains information of encoded objects
print("Point '%s' is similar to objects with indexes '%s'." % (str(point), str(index_similar_objects)))
print("Coordinates of similar objects:")
for index in index_similar_objects: print("\tPoint:", sample[index])

# result visualization:
# show distance matrix (U-matrix).
network.show_distance_matrix()

# show density matrix (P-matrix).
network.show_density_matrix()

# show winner matrix.
network.show_winner_matrix()

# show self-organized map.
network.show_network()

There is a visualization of 'Target' sample that was done by the self-organized feature map:

Definition at line 117 of file som.py.

Constructor & Destructor Documentation

__init__()

def pyclustering.nnet.som.som.__init__	(		self,
			rows,
			cols,
			conn_type = type_conn.grid_eight,
			parameters = None,
			ccore = True
	)

Constructor of self-organized map.

Parameters

[in]	rows	(uint): Number of neurons in the column (number of rows).
[in]	cols	(uint): Number of neurons in the row (number of columns).
[in]	conn_type	(type_conn): Type of connection between oscillators in the network (grid four, grid eight, honeycomb, function neighbour).
[in]	parameters	(som_parameters): Other specific parameters.
[in]	ccore	(bool): If True simulation is performed by CCORE library (C++ implementation of pyclustering).

Definition at line 247 of file som.py.

Member Function Documentation

__getstate__()

def pyclustering.nnet.som.som.__getstate__

(

self

)

@brief Returns state of SOM network that can be used to store network.

Definition at line 325 of file som.py.

__len__()

def pyclustering.nnet.som.som.__len__

(

self

)

Returns size of the network that defines by amount of neuron in it.

Returns

(uint) Size of self-organized map (amount of neurons).

Definition at line 315 of file som.py.

__setstate__()

def pyclustering.nnet.som.som.__setstate__	(		self,
			som_state
	)

@brief Set state of SOM network that can be used to load network.

Definition at line 336 of file som.py.

awards()

def pyclustering.nnet.som.som.awards

(

self

)

Return amount of captured objects by each neuron after training.

Returns

(list) Amount of captured objects by each neuron.

See also

train()

Definition at line 214 of file som.py.

capture_objects()

def pyclustering.nnet.som.som.capture_objects

(

self

)

Returns indexes of captured objects by each neuron.

For example, a network with size 2x2 has been trained on a sample with five objects. Suppose neuron #1 won an object with index 1, neuron #2 won objects 0, 3, 4, neuron #3 did not won anything and finally neuron #4 won an object with index 2. Thus, for this example we will have the following output [[1], [0, 3, 4], [], [2]].

Returns

(list) Indexes of captured objects by each neuron.

Definition at line 230 of file som.py.

get_density_matrix()

def pyclustering.nnet.som.som.get_density_matrix	(		self,
			surface_divider = 20.0
	)

Calculates density matrix (P-Matrix).

Parameters

[in]

surface_divider

(double): Divider in each dimension that affect radius for density measurement.

Returns

(list) Density matrix (P-Matrix).

See also

get_distance_matrix()

Definition at line 787 of file som.py.

Referenced by pyclustering.nnet.som.som.show_density_matrix().

get_distance_matrix()

def pyclustering.nnet.som.som.get_distance_matrix

(

self

)

Calculates distance matrix (U-matrix).

The U-Matrix visualizes based on the distance in input space between a weight vector and its neighbors on map.

Returns

(list) Distance matrix (U-matrix).

See also

show_distance_matrix()

get_density_matrix()

Definition at line 737 of file som.py.

Referenced by pyclustering.nnet.som.som.show_distance_matrix().

get_winner_number()

def pyclustering.nnet.som.som.get_winner_number

(

self

)

Calculates number of winner at the last step of learning process.

Returns

(uint) Number of winner.

Definition at line 705 of file som.py.

show_density_matrix()

def pyclustering.nnet.som.som.show_density_matrix	(		self,
			surface_divider = 20.0
	)

Show density matrix (P-matrix) using kernel density estimation.

Parameters

[in]

surface_divider

(double): Divider in each dimension that affect radius for density measurement.

See also

show_distance_matrix()

Definition at line 771 of file som.py.

show_distance_matrix()

def pyclustering.nnet.som.som.show_distance_matrix

(

self

)

Shows gray visualization of U-matrix (distance matrix).

See also

get_distance_matrix()

Definition at line 723 of file som.py.

show_network()

def pyclustering.nnet.som.som.show_network	(		self,
			awards = False,
			belongs = False,
			coupling = True,
			dataset = True,
			marker_type = 'o'
	)

Shows neurons in the dimension of data.

Parameters

[in]	awards	(bool): If True - displays how many objects won each neuron.
[in]	belongs	(bool): If True - marks each won object by according index of neuron-winner (only when dataset is displayed too).
[in]	coupling	(bool): If True - displays connections between neurons (except case when function neighbor is used).
[in]	dataset	(bool): If True - displays inputs data set.
[in]	marker_type	(string): Defines marker that is used to denote neurons on the plot.

Definition at line 866 of file som.py.

show_winner_matrix()

def pyclustering.nnet.som.som.show_winner_matrix

(

self

)

Show a winner matrix where each element corresponds to neuron and value represents amount of won objects from input data-space at the last training iteration.

See also

show_distance_matrix()

Definition at line 838 of file som.py.

simulate()

def pyclustering.nnet.som.som.simulate	(		self,
			input_pattern
	)

Processes input pattern (no learining) and returns index of neuron-winner.

Using index of neuron winner catched object can be obtained using property capture_objects.

Parameters

[in]

input_pattern

(list): Input pattern.

Returns

(uint) Returns index of neuron-winner.

See also

capture_objects

Definition at line 662 of file som.py.

size()

def pyclustering.nnet.som.som.size

(

self

)

Return size of self-organized map that is defined by total number of neurons.

Returns

(uint) Size of self-organized map (number of neurons).

Definition at line 186 of file som.py.

train()

def pyclustering.nnet.som.som.train	(		self,
			data,
			epochs,
			autostop = False
	)

Trains self-organized feature map (SOM).

Parameters

[in]	data	(list): Input data - list of points where each point is represented by list of features, for example coordinates.
[in]	epochs	(uint): Number of epochs for training.
[in]	autostop	(bool): Automatic termination of learning process when adaptation is not occurred.

Returns

(uint) Number of learning iterations.

Definition at line 600 of file som.py.

weights()

def pyclustering.nnet.som.som.weights

(

self

)

Return weight of each neuron.

Returns

(list) Weights of each neuron.

Definition at line 200 of file som.py.

---

The documentation for this class was generated from the following file:

• pyclustering/nnet/som.py