Python* API Reference for Intel® Data Analytics Acceleration Library 2018 Update 1

kmeans_dense_distr.py

1 #===============================================================================
2 # Copyright 2014-2017 Intel Corporation
3 # All Rights Reserved.
4 #
5 # If this software was obtained under the Intel Simplified Software License,
6 # the following terms apply:
7 #
8 # The source code, information and material ("Material") contained herein is
9 # owned by Intel Corporation or its suppliers or licensors, and title to such
10 # Material remains with Intel Corporation or its suppliers or licensors. The
11 # Material contains proprietary information of Intel or its suppliers and
12 # licensors. The Material is protected by worldwide copyright laws and treaty
13 # provisions. No part of the Material may be used, copied, reproduced,
14 # modified, published, uploaded, posted, transmitted, distributed or disclosed
15 # in any way without Intel's prior express written permission. No license under
16 # any patent, copyright or other intellectual property rights in the Material
17 # is granted to or conferred upon you, either expressly, by implication,
18 # inducement, estoppel or otherwise. Any license under such intellectual
19 # property rights must be express and approved by Intel in writing.
20 #
21 # Unless otherwise agreed by Intel in writing, you may not remove or alter this
22 # notice or any other notice embedded in Materials by Intel or Intel's
23 # suppliers or licensors in any way.
24 #
25 #
26 # If this software was obtained under the Apache License, Version 2.0 (the
27 # "License"), the following terms apply:
28 #
29 # You may not use this file except in compliance with the License. You may
30 # obtain a copy of the License at http://www.apache.org/licenses/LICENSE-2.0
31 #
32 #
33 # Unless required by applicable law or agreed to in writing, software
34 # distributed under the License is distributed on an "AS IS" BASIS, WITHOUT
35 # WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
36 #
37 # See the License for the specific language governing permissions and
38 # limitations under the License.
39 #===============================================================================
40 
41 
42 
43 
44 import os
45 import sys
46 
47 import daal.algorithms.kmeans as kmeans
48 import daal.algorithms.kmeans.init as init
49 from daal import step1Local, step2Master
50 from daal.data_management import FileDataSource, DataSourceIface
51 
52 utils_folder = os.path.realpath(os.path.abspath(os.path.dirname(os.path.dirname(__file__))))
53 if utils_folder not in sys.path:
54  sys.path.insert(0, utils_folder)
55 from utils import printNumericTable
56 
57 DAAL_PREFIX = os.path.join('..', 'data')
58 
59 dataFileNames = [
60  os.path.join(DAAL_PREFIX, 'distributed', 'kmeans_dense_1.csv'),
61  os.path.join(DAAL_PREFIX, 'distributed', 'kmeans_dense_2.csv'),
62  os.path.join(DAAL_PREFIX, 'distributed', 'kmeans_dense_3.csv'),
63  os.path.join(DAAL_PREFIX, 'distributed', 'kmeans_dense_4.csv')
64 ]
65 
66 nClusters = 20
67 nIterations = 5
68 nBlocks = 4
69 nVectorsInBlock = 2500
70 
71 dataTable = [0] * nBlocks
72 
73 if __name__ == "__main__":
74 
75  masterAlgorithm = kmeans.Distributed(step2Master, nClusters, method=kmeans.lloydDense)
76 
77  centroids = None
78  assignments = [0] * nBlocks
79 
80  masterInitAlgorithm = init.Distributed(step2Master, nClusters, method=init.randomDense)
81  for i in range(nBlocks):
82  # Initialize FileDataSource<CSVFeatureManager> to retrieve the input data from a .csv file
83  dataSource = FileDataSource(
84  dataFileNames[i], DataSourceIface.doAllocateNumericTable,
85  DataSourceIface.doDictionaryFromContext
86  )
87 
88  # Retrieve the data from the input file
89  dataSource.loadDataBlock()
90 
91  dataTable[i] = dataSource.getNumericTable()
92 
93  # Create an algorithm object for the K-Means algorithm
94  localInit = init.Distributed(step1Local, nClusters, nBlocks * nVectorsInBlock, i * nVectorsInBlock, method=init.randomDense)
95 
96  localInit.input.set(init.data, dataTable[i])
97  res = localInit.compute()
98  masterInitAlgorithm.input.add(init.partialResults, res)
99 
100  masterInitAlgorithm.compute()
101  res = masterInitAlgorithm.finalizeCompute()
102  centroids = res.get(init.centroids)
103 
104  for it in range(nIterations):
105  for i in range(nBlocks):
106  # Create an algorithm object for the K-Means algorithm
107  localAlgorithm = kmeans.Distributed(step1Local, nClusters, it == nIterations, method=kmeans.lloydDense)
108 
109  # Set the input data to the algorithm
110  localAlgorithm.input.set(kmeans.data, dataTable[i])
111  localAlgorithm.input.set(kmeans.inputCentroids, centroids)
112 
113  pres = localAlgorithm.compute()
114 
115  masterAlgorithm.input.add(kmeans.partialResults, pres)
116 
117  masterAlgorithm.compute()
118  result = masterAlgorithm.finalizeCompute()
119 
120  centroids = result.get(kmeans.centroids)
121  goalFunction = result.get(kmeans.goalFunction)
122 
123  for i in range(nBlocks):
124  # Create an algorithm object for the K-Means algorithm
125  localAlgorithm = kmeans.Batch(nClusters, 0, method=kmeans.lloydDense)
126 
127  # Set the input data to the algorithm
128  localAlgorithm.input.set(kmeans.data, dataTable[i])
129  localAlgorithm.input.set(kmeans.inputCentroids, centroids)
130 
131  res = localAlgorithm.compute()
132 
133  assignments[i] = res.get(kmeans.assignments)
134 
135  # Print the clusterization results
136  printNumericTable(assignments[0], "First 10 cluster assignments from 1st node:", 10)
137  printNumericTable(centroids, "First 10 dimensions of centroids:", 20, 10)
138  printNumericTable(goalFunction, "Goal function value:")

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