scripts/utils/sample.dml

#-------------------------------------------------------------
#
# Licensed to the Apache Software Foundation (ASF) under one
# or more contributor license agreements.  See the NOTICE file
# distributed with this work for additional information
# regarding copyright ownership.  The ASF licenses this file
# to you under the Apache License, Version 2.0 (the
# "License"); you may not use this file except in compliance
# with the License.  You may obtain a copy of the License at
# 
#   http://www.apache.org/licenses/LICENSE-2.0
# 
# Unless required by applicable law or agreed to in writing,
# software distributed under the License is distributed on an
# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
# KIND, either express or implied.  See the License for the
# specific language governing permissions and limitations
# under the License.
#
#-------------------------------------------------------------

# Randomly sample data (without replacement) into disjoint subsets.
# The sizes of the subsets are specified in terms of fractions, stored
# as a 1-column vector in a separate input file (see parameter 'sv'). 
#
# Parameters:
#    X    : (input) input data set: filename of input data set
#    sv   : (input) splitting vector: filename of 1-column vector with
#           fractions. sum(sv) must be less than or equal to 1
#               e.g. sv = [0.2]: Draw a 20% simple random sample
#                    without replacement.
#               e.g. sv = [0.25,0.25,0.25,0.25]: Randomly split data
#                    into 4 approximately equal-sized disjoint subsets.
#               e.g. sv = [0.5,0.3,0.2]: Randomly split data into 3
#                    disjoint subsets that contain roughly 50%, 30%
#                    and 20% of original data, respectively.
#    O    : (output) output folder name. The output subsets are stored
#           in subfolders named by consecutive integers: $O/1, $O/2,
#           ..., $O/#subsets
#    ofmt : (output, default "binary") format of output file. Other
#           valid options are: "csv" and "text" 
#
# Example:
#   printf "0.8\n0.2" | hadoop fs -put - /tmp/sv.csv
#   echo '{"data_type": "matrix", "value_type": "double", "rows": 2, "cols": 1, "format": "csv"}' | hadoop fs -put - /tmp/sv.csv.mtd
#   hadoop jar SystemML.jar -f ./scripts/utils/sample.dml -nvargs X=/tmp/X.mtx sv=/tmp/sv.csv O=/tmp/Out ofmt=csv

# set defaults
ofmt = ifdef($ofmt, "binary");

# Read inputs
X = read ($X);         # X: dataset
sv = read ($sv);       # sv: splitting fraction vector

# Construct sampling lower/upper bounds for samples using prefix sum
R = rand(rows=nrow(X), cols=1, min=0.0, max=1.0, pdf = "uniform");
svLowBnd = cumsum(sv) - sv;
svUpBnd = cumsum(sv);

# Construct sampling matrix SM, and apply to create samples
parfor ( i in 1:nrow(sv))
{
  T1 = R <= as.scalar(svUpBnd[i,1]);
  T2 = R > as.scalar(svLowBnd[i,1]);
  SM = T1 * T2; 
  P = removeEmpty(target=diag(SM), margin="rows");
  iX = P %*% X;
  write (iX, $O + "/" + i, format=ofmt);
}