kernelpca

import math as m
import sys
import numpy as n
from numpy import *
import matplotlib.pyplot as p
#from tools.load import LoadMatrix

d=[[0 for x in xrange(101)] for x in xrange(2)]
for x in range(-50,1):
    d[0][x+50]=x
for x in range(1,51):
    d[0][x+50]=x

i=0
while(i<51):
        #print i
    y2=2500-(i*i)
    yy2=m.sqrt(y2)
    d[1][i+50]=yy2
    i=i+1
i=50
while(i>0):
        #print
    y2=2500-(i*i)
    yy2=m.sqrt(y2)
    d[1][50-i]=yy2
    i=i-1
        #d[1][100+x]=yy2



    #print d[0][:],'\n',d[1][:]
    #p.plot(d[0][:],d[1][:],'x')

d2=[[0 for x in xrange(201)] for x in xrange(2)]
for x in range(-100,1):
    d2[0][x+100]=x
for x in range(1,101):
    d2[0][x+100]=x

i=0
while(i<101):
        #print i
    y2=10000-(i*i)
    yy2=m.sqrt(y2)
    d2[1][i+100]=yy2
    i=i+1
i=100
while(i>0):
        #print
    y2=10000-(i*i)
    yy2=m.sqrt(y2)
    d2[1][100-i]=yy2
    i=i-1
        #d[1][100+x]=yy2
    #print d2[0][:],'\n',d2[1][:]
p.plot(d[0][:],d[1][:],'x',d2[0][:],d2[1][:],'o')
data=hstack((d,d2))
#data=d
print 'size =',n.shape(data)
p.show()

parameter_list = [[data,0.01,1.0], [data,0.05,2.0]]

def preprocessor_kernelpca_modular (data, threshold, width):
    from shogun.Features import RealFeatures
    from shogun.Preprocessor import KernelPCA
    from shogun.Kernel import GaussianKernel
    #from shogun.Converter import StochasticProximityEmbedding, SPE_GLOBAL
    features = RealFeatures(data)

    kernel=GaussianKernel(features,features,width)

    preprocessor=KernelPCA(kernel)
    preprocessor.init(features)
    preprocessor.apply_to_feature_matrix(features)
    X = preprocessor.get_transformation_matrix()
    #l1=len(X)

    print 'the rows=%d, ',n.shape(X)
    p.plot(X[0][:],'x',X[1][:],'o')
    p.show()
    print 'type of features=%',(type(X))
    return features


if __name__=='__main__':
    print('KernelPCA')
    preprocessor_kernelpca_modular(*parameter_list[0])