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

ridge_reg_norm_eq_dense_distr.py

1 # file: ridge_reg_norm_eq_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 #
43 # ! Content:
44 # ! Python example of ridge regression in the distributed processing mode.
45 # !
46 # ! The program trains the multiple ridge regression model on a training
47 # ! datasetFileName with the normal equations method and computes regression
48 # ! for the test data.
49 # !*****************************************************************************
50 
51 #
52 
55 
56 import os
57 import sys
58 
59 from daal import step1Local, step2Master
60 from daal.algorithms.ridge_regression import training, prediction
61 from daal.data_management import DataSource, FileDataSource, NumericTable, HomogenNumericTable, MergedNumericTable
62 
63 utils_folder = os.path.realpath(os.path.abspath(os.path.dirname(os.path.dirname(__file__))))
64 if utils_folder not in sys.path:
65  sys.path.insert(0, utils_folder)
66 from utils import printNumericTable
67 
68 trainDatasetFileNames = [
69  os.path.join("..", "data", "distributed", "linear_regression_train_1.csv"),
70  os.path.join("..", "data", "distributed", "linear_regression_train_2.csv"),
71  os.path.join("..", "data", "distributed", "linear_regression_train_3.csv"),
72  os.path.join("..", "data", "distributed", "linear_regression_train_4.csv"),
73 ]
74 
75 
76 testDatasetFileName = os.path.join("..", "data", "distributed", "linear_regression_test.csv")
77 
78 nBlocks = 4
79 nFeatures = 10 # Number of features in training and testing data sets
80 nDependentVariables = 2 # Number of dependent variables that correspond to each observation
81 
82 
83 def trainModel():
84  # Create an algorithm object to build the final ridge regression model on the master node
85  masterAlgorithm = training.Distributed(step=step2Master)
86 
87  for i in range(nBlocks):
88  # Initialize FileDataSource<CSVFeatureManager> to retrieve the input data from a .csv file
89  trainDataSource = FileDataSource(trainDatasetFileNames[i],
90  DataSource.notAllocateNumericTable,
91  DataSource.doDictionaryFromContext)
92 
93  # Create Numeric Tables for training data and variables
94  trainData = HomogenNumericTable(nFeatures, 0, NumericTable.doNotAllocate)
95  trainDependentVariables = HomogenNumericTable(nDependentVariables, 0, NumericTable.doNotAllocate)
96  mergedData = MergedNumericTable(trainData, trainDependentVariables)
97 
98  # Retrieve the data from input file
99  trainDataSource.loadDataBlock(mergedData)
100 
101  # Create an algorithm object to train the ridge regression model based on the local-node data
102  localAlgorithm = training.Distributed(step=step1Local)
103 
104  # Pass a training data set and dependent values to the algorithm
105  localAlgorithm.input.set(training.data, trainData)
106  localAlgorithm.input.set(training.dependentVariables, trainDependentVariables)
107 
108  # Train the ridge regression model on the local-node data
109  presult = localAlgorithm.compute()
110 
111  # Set the local ridge regression model as input for the master-node algorithm
112  masterAlgorithm.input.add(training.partialModels, presult)
113 
114 
115  # Merge and finalize the ridge regression model on the master node
116  masterAlgorithm.compute()
117 
118  # Retrieve the algorithm results
119  trainingResult = masterAlgorithm.finalizeCompute()
120 
121  printNumericTable(trainingResult.get(training.model).getBeta(), "Ridge Regression coefficients:")
122  return trainingResult
123 
124 
125 def testModel(trainingResult):
126  # Initialize FileDataSource<CSVFeatureManager> to retrieve the input data from a .csv file
127  testDataSource = FileDataSource(testDatasetFileName,
128  DataSource.doAllocateNumericTable,
129  DataSource.doDictionaryFromContext)
130 
131  # Create Numeric Tables for testing data and ground truth values
132  testData = HomogenNumericTable(nFeatures, 0, NumericTable.doNotAllocate)
133  testGroundTruth = HomogenNumericTable(nDependentVariables, 0, NumericTable.doNotAllocate)
134  mergedData = MergedNumericTable(testData, testGroundTruth)
135 
136  # Load the data from the data file
137  testDataSource.loadDataBlock(mergedData)
138 
139  # Create an algorithm object to predict values of ridge regression
140  algorithm = prediction.Batch()
141 
142  # Pass a testing data set and the trained model to the algorithm
143  algorithm.input.setTable(prediction.data, testData)
144  algorithm.input.setModel(prediction.model, trainingResult.get(training.model))
145 
146  # Predict values of ridge regression and retrieve the algorithm results
147  predictionResult = algorithm.compute()
148 
149  printNumericTable(predictionResult.get(prediction.prediction), "Ridge Regression prediction results: (first 10 rows):", 10)
150  printNumericTable(testGroundTruth, "Ground truth (first 10 rows):", 10)
151 
152 
153 if __name__ == "__main__":
154  trainingResult = trainModel()
155  testModel(trainingResult)

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