[mlpack-svn] r15418 - mlpack/conf/jenkins-conf/benchmark/methods/matlab
fastlab-svn at coffeetalk-1.cc.gatech.edu
fastlab-svn at coffeetalk-1.cc.gatech.edu
Fri Jul 5 12:57:20 EDT 2013
Author: marcus
Date: Fri Jul 5 12:57:20 2013
New Revision: 15418
Log:
Add matlab K-Means method and K-Means benchmark script.
Added:
mlpack/conf/jenkins-conf/benchmark/methods/matlab/KMEANS.m
mlpack/conf/jenkins-conf/benchmark/methods/matlab/kmeans.py
Added: mlpack/conf/jenkins-conf/benchmark/methods/matlab/KMEANS.m
==============================================================================
--- (empty file)
+++ mlpack/conf/jenkins-conf/benchmark/methods/matlab/KMEANS.m Fri Jul 5 12:57:20 2013
@@ -0,0 +1,61 @@
+% @file KMEANS.m
+% @author Marcus Edel
+%
+% K-Means Clustering with matlab.
+
+function KMEANS(cmd)
+% This program performs K-Means clustering on the given dataset
+%
+% Required options:
+% (-i) [string] Input dataset to perform clustering on.
+% Options:
+% (-c) [int] Number of clusters to find.
+% (-m) [int] Maximum number of iterations before K-Means
+% terminates. Default value 1000.
+% (-s) [int] Random seed. If 0, 'std::time(NULL)' is used.
+
+
+% Load input dataset.
+inputFile = regexp(cmd, '.*?-i ([^\s]+)', 'tokens', 'once');
+X = csvread(inputFile{:});
+
+% Check if centroid starting locations set is given.
+C = [];
+if strfind(cmd, '-I') > 0
+ centroidFile = regexp(cmd, '.*?-I ([^\s]+)', 'tokens', 'once');
+ C = csvread(centroidFile{:});
+end
+
+% Gather parameters.
+clusters = str2double(regexp(cmd,'.* -c (\d+)','tokens','once'));
+maxIterations = str2double(regexp(cmd,'.* -m (\d+)','tokens','once'));
+seed = str2double(regexp(cmd,'.* -s (\d+)','tokens','once'));
+
+% Validate parameters.
+if isempty(maxIterations)
+ m = 1000;
+else
+ if maxIterations == 0
+ m = inf;
+ elseif maxIterations
+ m = maxIterations;
+ end
+end
+
+if ~isempty(seed)
+ s = RandStream('mt19937ar','Seed', seed);
+ RandStream.setGlobalStream(s);
+end
+
+total_time = tic;
+if ~isempty(clusters)
+ [IDX, C] = kmeans(X, clusters, 'EmptyAction', 'singleton', ...
+ 'MaxIter', m);
+ disp(sprintf('[INFO ] total_time: %fs', toc(total_time)))
+elseif ~isempty(C)
+ [IDX, C] = kmeans(X, size(C, 1), 'Start', C, 'EmptyAction', ...
+ 'singleton', 'MaxIter', m);
+ disp(sprintf('[INFO ] total_time: %fs', toc(total_time)))
+end
+
+end
Added: mlpack/conf/jenkins-conf/benchmark/methods/matlab/kmeans.py
==============================================================================
--- (empty file)
+++ mlpack/conf/jenkins-conf/benchmark/methods/matlab/kmeans.py Fri Jul 5 12:57:20 2013
@@ -0,0 +1,121 @@
+'''
+ @file kmeans.py
+ @author Marcus Edel
+
+ Class to benchmark the matlab K-Means Clustering method.
+'''
+
+import os
+import sys
+import inspect
+
+# Import the util path, this method even works if the path contains symlinks to
+# modules.
+cmd_subfolder = os.path.realpath(os.path.abspath(os.path.join(
+ os.path.split(inspect.getfile(inspect.currentframe()))[0], "../../util")))
+if cmd_subfolder not in sys.path:
+ sys.path.insert(0, cmd_subfolder)
+
+from log import *
+from profiler import *
+
+import shlex
+import subprocess
+import re
+import collections
+
+'''
+This class implements the K-Means Clustering benchmark.
+'''
+class KMEANS(object):
+
+ '''
+ Create the K-Means Clustering benchmark instance.
+
+ @param dataset - Input dataset to perform K-Means on.
+ @param path - Path to the mlpack executable.
+ @param verbose - Display informational messages.
+ '''
+ def __init__(self, dataset, path=os.environ["MATLAB_BIN"], verbose = True):
+ self.verbose = verbose
+ self.dataset = dataset
+ self.path = path
+
+ '''
+ Destructor to clean up at the end.
+ '''
+ def __del__(self):
+ pass
+
+ '''
+ Non-negative Matrix Factorization. If the method has been successfully
+ completed return the elapsed time in seconds.
+
+ @param options - Extra options for the method.
+ @return - Elapsed time in seconds or -1 if the method was not successful.
+ '''
+ def RunMethod(self, options):
+ Log.Info("Perform K-Means.", self.verbose)
+
+ # If the dataset contains two files then the second file is the centroids
+ # file. In this case we add this to the command line.
+ if len(self.dataset) == 2:
+ inputCmd = "-i " + self.dataset[0] + " -I " + self.dataset[1] + " " + options
+ else:
+ inputCmd = "-i " + self.dataset + " " + options
+
+ # Split the command using shell-like syntax.
+ cmd = shlex.split(self.path + "matlab -nodisplay -nosplash -r \"try, " +
+ "KMEANS('" + inputCmd + "'), catch, exit(1), end, exit(0)\"")
+
+ # Run command with the nessecary arguments and return its output as a byte
+ # string. We have untrusted input so we disables all shell based features.
+ try:
+ s = subprocess.check_output(cmd, stderr=subprocess.STDOUT, shell=False)
+ except Exception:
+ Log.Fatal("Could not execute command: " + str(cmd))
+ return -1
+
+ # Return the elapsed time.
+ timer = self.parseTimer(s)
+ if not timer:
+ Log.Fatal("Can't parse the timer")
+ return -1
+ else:
+ time = self.GetTime(timer)
+ Log.Info(("total time: %fs" % time), self.verbose)
+
+ return time
+
+ '''
+ Parse the timer data form a given string.
+
+ @param data - String to parse timer data from.
+ @return - Namedtuple that contains the timer data.
+ '''
+ def parseTimer(self, data):
+ # Compile the regular expression pattern into a regular expression object to
+ # parse the timer data.
+ pattern = re.compile(r"""
+ .*?total_time: (?P<total_time>.*?)s.*?
+ """, re.VERBOSE|re.MULTILINE|re.DOTALL)
+
+ match = pattern.match(data)
+ if not match:
+ Log.Fatal("Can't parse the data: wrong format")
+ return -1
+ else:
+ # Create a namedtuple and return the timer data.
+ timer = collections.namedtuple("timer", ["total_time"])
+
+ return timer(float(match.group("total_time")))
+
+ '''
+ Return the elapsed time in seconds.
+
+ @param timer - Namedtuple that contains the timer data.
+ @return Elapsed time in seconds.
+ '''
+ def GetTime(self, timer):
+ time = timer.total_time
+ return time
More information about the mlpack-svn
mailing list