Mercurial > repos > devteam > kernel_principal_component_analysis

comparison kpca.py @ 0:e9ebd4bfbdfc draft

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
Imported from capsule None
author devteam
date Mon, 19 May 2014 12:34:31 -0400
parents
children
comparison
equal deleted inserted replaced
-1:000000000000 0:e9ebd4bfbdfc
1 #!/usr/bin/env python
2
3 """
4 Run kernel PCA using kpca() from R 'kernlab' package
5
6 usage: %prog [options]
7 -i, --input=i: Input file
8 -o, --output1=o: Summary output
9 -p, --output2=p: Figures output
10 -c, --var_cols=c: Variable columns
11 -k, --kernel=k: Kernel function
12 -f, --features=f: Number of principal components to return
13 -s, --sigma=s: sigma
14 -d, --degree=d: degree
15 -l, --scale=l: scale
16 -t, --offset=t: offset
17 -r, --order=r: order
18
19 usage: %prog input output1 output2 var_cols kernel features sigma(or_None) degree(or_None) scale(or_None) offset(or_None) order(or_None)
20 """
21
22 import sys, string
23 from rpy import *
24 import numpy
25 from bx.cookbook import doc_optparse
26
27
28 def stop_err(msg):
29 sys.stderr.write(msg)
30 sys.exit()
31
32 #Parse Command Line
33 options, args = doc_optparse.parse( __doc__ )
34 #{'options= kernel': 'rbfdot', 'var_cols': '1,2,3,4', 'degree': 'None', 'output2': '/afs/bx.psu.edu/home/gua110/workspace/galaxy_bitbucket/database/files/000/dataset_260.dat', 'output1': '/afs/bx.psu.edu/home/gua110/workspace/galaxy_bitbucket/database/files/000/dataset_259.dat', 'scale': 'None', 'offset': 'None', 'input': '/afs/bx.psu.edu/home/gua110/workspace/galaxy_bitbucket/database/files/000/dataset_256.dat', 'sigma': '1.0', 'order': 'None'}
35
36 infile = options.input
37 x_cols = options.var_cols.split(',')
38 kernel = options.kernel
39 outfile = options.output1
40 outfile2 = options.output2
41 ncomps = int(options.features)
42 fout = open(outfile,'w')
43
44 elems = []
45 for i, line in enumerate( file ( infile )):
46 line = line.rstrip('\r\n')
47 if len( line )>0 and not line.startswith( '#' ):
48 elems = line.split( '\t' )
49 break
50 if i == 30:
51 break # Hopefully we'll never get here...
52
53 if len( elems )<1:
54 stop_err( "The data in your input dataset is either missing or not formatted properly." )
55
56 x_vals = []
57
58 for k,col in enumerate(x_cols):
59 x_cols[k] = int(col)-1
60 x_vals.append([])
61
62 NA = 'NA'
63 skipped = 0
64 for ind,line in enumerate( file( infile )):
65 if line and not line.startswith( '#' ):
66 try:
67 fields = line.strip().split("\t")
68 for k,col in enumerate(x_cols):
69 try:
70 xval = float(fields[col])
71 except:
72 #xval = r('NA')
73 xval = NaN#
74 x_vals[k].append(xval)
75 except:
76 skipped += 1
77
78 x_vals1 = numpy.asarray(x_vals).transpose()
79 dat= r.list(array(x_vals1))
80
81 print r('library("kernlab")')
82
83 try:
84 r.suppressWarnings(r.library('kernlab'))
85 except:
86 stop_err('Missing R library kernlab')
87
88 set_default_mode(NO_CONVERSION)
89 if kernel=="rbfdot" or kernel=="anovadot":
90 pars = r.list(sigma=float(options.sigma))
91 elif kernel=="polydot":
92 pars = r.list(degree=float(options.degree),scale=float(options.scale),offset=float(options.offset))
93 elif kernel=="tanhdot":
94 pars = r.list(scale=float(options.scale),offset=float(options.offset))
95 elif kernel=="besseldot":
96 pars = r.list(degree=float(options.degree),sigma=float(options.sigma),order=float(options.order))
97 elif kernel=="anovadot":
98 pars = r.list(degree=float(options.degree),sigma=float(options.sigma))
99 else:
100 pars = r.list()
101
102 try:
103 kpc = r.kpca(x=r.na_exclude(dat), kernel=kernel, kpar=pars, features=ncomps)
104 except RException, rex:
105 stop_err("Encountered error while performing kPCA on the input data: %s" %(rex))
106 set_default_mode(BASIC_CONVERSION)
107
108 eig = r.eig(kpc)
109 pcv = r.pcv(kpc)
110 rotated = r.rotated(kpc)
111
112 comps = eig.keys()
113 eigv = eig.values()
114 for i in range(ncomps):
115 eigv[comps.index('Comp.%s' %(i+1))] = eig.values()[i]
116
117 print >>fout, "#Component\t%s" %("\t".join(["%s" % el for el in range(1,ncomps+1)]))
118
119 print >>fout, "#Eigenvalue\t%s" %("\t".join(["%.4g" % el for el in eig.values()]))
120
121 print >>fout, "#Principal component vectors\t%s" %("\t".join(["%s" % el for el in range(1,ncomps+1)]))
122 for obs,val in enumerate(pcv):
123 print >>fout, "%s\t%s" %(obs+1, "\t".join(["%.4g" % el for el in val]))
124
125 print >>fout, "#Rotated values\t%s" %("\t".join(["%s" % el for el in range(1,ncomps+1)]))
126 for obs,val in enumerate(rotated):
127 print >>fout, "%s\t%s" %(obs+1, "\t".join(["%.4g" % el for el in val]))
128
129 r.pdf( outfile2, 8, 8 )
130 if ncomps != 1:
131 r.pairs(rotated,labels=r.list(range(1,ncomps+1)),main="Scatterplot of rotated values")
132 else:
133 r.plot(rotated, ylab='Comp.1', main="Scatterplot of rotated values")
134 r.dev_off()