Python* API Reference for Intel® Data Analytics Acceleration Library 2018 Update 2

kmeans_dense_distr.py

1 # file: kmeans_dense_distr.py
2 #===============================================================================
3 # Copyright 2014-2018 Intel Corporation
4 # All Rights Reserved.
5 #
6 # If this software was obtained under the Intel Simplified Software License,
7 # the following terms apply:
8 #
9 # The source code, information and material ("Material") contained herein is
10 # owned by Intel Corporation or its suppliers or licensors, and title to such
11 # Material remains with Intel Corporation or its suppliers or licensors. The
12 # Material contains proprietary information of Intel or its suppliers and
13 # licensors. The Material is protected by worldwide copyright laws and treaty
14 # provisions. No part of the Material may be used, copied, reproduced,
15 # modified, published, uploaded, posted, transmitted, distributed or disclosed
16 # in any way without Intel's prior express written permission. No license under
17 # any patent, copyright or other intellectual property rights in the Material
18 # is granted to or conferred upon you, either expressly, by implication,
19 # inducement, estoppel or otherwise. Any license under such intellectual
20 # property rights must be express and approved by Intel in writing.
21 #
22 # Unless otherwise agreed by Intel in writing, you may not remove or alter this
23 # notice or any other notice embedded in Materials by Intel or Intel's
24 # suppliers or licensors in any way.
25 #
26 #
27 # If this software was obtained under the Apache License, Version 2.0 (the
28 # "License"), the following terms apply:
29 #
30 # You may not use this file except in compliance with the License. You may
31 # obtain a copy of the License at http://www.apache.org/licenses/LICENSE-2.0
32 #
33 #
34 # Unless required by applicable law or agreed to in writing, software
35 # distributed under the License is distributed on an "AS IS" BASIS, WITHOUT
36 # WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
37 #
38 # See the License for the specific language governing permissions and
39 # limitations under the License.
40 #===============================================================================
41 
42 
44 
45 import os
46 import sys
47 
48 import daal.algorithms.kmeans as kmeans
49 import daal.algorithms.kmeans.init as init
50 from daal import step1Local, step2Master
51 from daal.data_management import FileDataSource, DataSourceIface
52 
53 utils_folder = os.path.realpath(os.path.abspath(os.path.dirname(os.path.dirname(__file__))))
54 if utils_folder not in sys.path:
55  sys.path.insert(0, utils_folder)
56 from utils import printNumericTable
57 
58 DAAL_PREFIX = os.path.join('..', 'data')
59 
60 dataFileNames = [
61  os.path.join(DAAL_PREFIX, 'distributed', 'kmeans_dense_1.csv'),
62  os.path.join(DAAL_PREFIX, 'distributed', 'kmeans_dense_2.csv'),
63  os.path.join(DAAL_PREFIX, 'distributed', 'kmeans_dense_3.csv'),
64  os.path.join(DAAL_PREFIX, 'distributed', 'kmeans_dense_4.csv')
65 ]
66 
67 nClusters = 20
68 nIterations = 5
69 nBlocks = 4
70 nVectorsInBlock = 2500
71 
72 dataTable = [0] * nBlocks
73 
74 if __name__ == "__main__":
75 
76  masterAlgorithm = kmeans.Distributed(step2Master, nClusters, method=kmeans.lloydDense)
77 
78  centroids = None
79  assignments = [0] * nBlocks
80 
81  masterInitAlgorithm = init.Distributed(step2Master, nClusters, method=init.randomDense)
82  for i in range(nBlocks):
83  # Initialize FileDataSource<CSVFeatureManager> to retrieve the input data from a .csv file
84  dataSource = FileDataSource(
85  dataFileNames[i], DataSourceIface.doAllocateNumericTable,
86  DataSourceIface.doDictionaryFromContext
87  )
88 
89  # Retrieve the data from the input file
90  dataSource.loadDataBlock()
91 
92  dataTable[i] = dataSource.getNumericTable()
93 
94  # Create an algorithm object for the K-Means algorithm
95  localInit = init.Distributed(step1Local, nClusters, nBlocks * nVectorsInBlock, i * nVectorsInBlock, method=init.randomDense)
96 
97  localInit.input.set(init.data, dataTable[i])
98  res = localInit.compute()
99  masterInitAlgorithm.input.add(init.partialResults, res)
100 
101  masterInitAlgorithm.compute()
102  res = masterInitAlgorithm.finalizeCompute()
103  centroids = res.get(init.centroids)
104 
105  for it in range(nIterations):
106  for i in range(nBlocks):
107  # Create an algorithm object for the K-Means algorithm
108  localAlgorithm = kmeans.Distributed(step1Local, nClusters, it == nIterations, method=kmeans.lloydDense)
109 
110  # Set the input data to the algorithm
111  localAlgorithm.input.set(kmeans.data, dataTable[i])
112  localAlgorithm.input.set(kmeans.inputCentroids, centroids)
113 
114  pres = localAlgorithm.compute()
115 
116  masterAlgorithm.input.add(kmeans.partialResults, pres)
117 
118  masterAlgorithm.compute()
119  result = masterAlgorithm.finalizeCompute()
120 
121  centroids = result.get(kmeans.centroids)
122  goalFunction = result.get(kmeans.goalFunction)
123 
124  for i in range(nBlocks):
125  # Create an algorithm object for the K-Means algorithm
126  localAlgorithm = kmeans.Batch(nClusters, 0, method=kmeans.lloydDense)
127 
128  # Set the input data to the algorithm
129  localAlgorithm.input.set(kmeans.data, dataTable[i])
130  localAlgorithm.input.set(kmeans.inputCentroids, centroids)
131 
132  res = localAlgorithm.compute()
133 
134  assignments[i] = res.get(kmeans.assignments)
135 
136  # Print the clusterization results
137  printNumericTable(assignments[0], "First 10 cluster assignments from 1st node:", 10)
138  printNumericTable(centroids, "First 10 dimensions of centroids:", 20, 10)
139  printNumericTable(goalFunction, "Goal function value:")

For more complete information about compiler optimizations, see our Optimization Notice.