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

kmeans_csr_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 ## <a name="DAAL-EXAMPLE-PY-KMEANS_CSR_DISTRIBUTED"></a>
42 ## \example kmeans_csr_distr.py
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 
51 utils_folder = os.path.realpath(os.path.abspath(os.path.dirname(os.path.dirname(__file__))))
52 if utils_folder not in sys.path:
53  sys.path.insert(0, utils_folder)
54 from utils import printNumericTable, createSparseTable
55 
56 DAAL_PREFIX = os.path.join('..', 'data')
57 
58 # K-Means algorithm parameters
59 nClusters = 20
60 nIterations = 5
61 nBlocks = 4
62 nVectorsInBlock = 8000
63 
64 dataFileNames = [
65  os.path.join(DAAL_PREFIX, 'batch', 'kmeans_csr.csv'),
66  os.path.join(DAAL_PREFIX, 'batch', 'kmeans_csr.csv'),
67  os.path.join(DAAL_PREFIX, 'batch', 'kmeans_csr.csv'),
68  os.path.join(DAAL_PREFIX, 'batch', 'kmeans_csr.csv')
69 ]
70 
71 dataTable = [0] * nBlocks
72 
73 if __name__ == "__main__":
74 
75  masterAlgorithm = kmeans.Distributed(step2Master, nClusters, method=kmeans.lloydCSR, )
76 
77  centroids = None
78  assignments = [0] * nBlocks
79 
80  masterInitAlgorithm = init.Distributed(step2Master, nClusters, method=init.randomDense)
81 
82  for i in range(nBlocks):
83 
84  # Read dataFileNames and create a numeric table to store the input data
85  dataTable[i] = createSparseTable(dataFileNames[i])
86 
87  # Create an algorithm object for the K-Means algorithm
88  localInit = init.Distributed(step1Local, nClusters, nBlocks * nVectorsInBlock, i * nVectorsInBlock, method=init.randomDense)
89 
90  localInit.input.set(init.data, dataTable[i])
91  # compute and add input for next
92  masterInitAlgorithm.input.add(init.partialResults, localInit.compute())
93 
94  masterInitAlgorithm.compute()
95  res = masterInitAlgorithm.finalizeCompute()
96  centroids = res.get(init.centroids)
97 
98  for it in range(nIterations):
99  for i in range(nBlocks):
100  # Create an algorithm object for the K-Means algorithm
101  localAlgorithm = kmeans.Distributed(step1Local, nClusters, it == nIterations, method=kmeans.lloydCSR)
102 
103  # Set the input data to the algorithm
104  localAlgorithm.input.set(kmeans.data, dataTable[i])
105  localAlgorithm.input.set(kmeans.inputCentroids, centroids)
106 
107  pres = localAlgorithm.compute()
108 
109  masterAlgorithm.input.add(kmeans.partialResults, pres)
110 
111  masterAlgorithm.compute()
112  result = masterAlgorithm.finalizeCompute()
113 
114  centroids = result.get(kmeans.centroids)
115  objectiveFunction = result.get(kmeans.objectiveFunction)
116 
117  for i in range(nBlocks):
118  # Create an algorithm object for the K-Means algorithm
119  localAlgorithm = kmeans.Batch(nClusters, 0, method=kmeans.lloydCSR)
120 
121  # Set the input data to the algorithm
122  localAlgorithm.input.set(kmeans.data, dataTable[i])
123  localAlgorithm.input.set(kmeans.inputCentroids, centroids)
124 
125  res = localAlgorithm.compute()
126 
127  assignments[i] = res.get(kmeans.assignments)
128 
129  # Print the clusterization results
130  printNumericTable(assignments[0], "First 10 cluster assignments from 1st node:", 10)
131  printNumericTable(centroids, "First 10 dimensions of centroids:", 20, 10)
132  printNumericTable(objectiveFunction, "Objective function value:")

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