def pima(self, dims=[1, 5, 6, 7], depth=4, width=2):
data = pods.datasets.pima(dims=dims)
X, Y, XLabel = data['X'], data['Y'], data['XLabel']
d = GPCData(X, Y, XLabel=XLabel, YLabel=['not diabetic', 'diabetic'])
print "\n\nPima Indian Diabetes dataset"
print "=====\nData size: D = %d, N = %d." % (d.getDim(), d.getNum())
search = GPCSearch(data=d, max_depth=depth, beam_width=width)
best, best1d = search.search()
constker = search.baseline()
report = GPCReport(name='Pima',
history=best,
best1d=best1d,
constkernel=constker)
report.export()
print ""
def bupa(self, dims=range(5), depth=4, width=2):
data = pods.datasets.bupa(dims=dims)
X, Y, XLabel = data['X'], data['Y'], data['XLabel']
d = GPCData(X,
Y,
XLabel=XLabel,
YLabel=['$\\leq 5$ drink units', '$> 5$ drink units'])
print "\n\nBUPA Liver Disorders dataset"
print "=====\nData size: D = %d, N = %d." % (d.getDim(), d.getNum())
search = GPCSearch(data=d, max_depth=depth, beam_width=width)
best, best1d = search.search()
constker = search.baseline()
report = GPCReport(name='BUPA',
history=best,
best1d=best1d,
constkernel=constker)
report.export()
def wisconsin(self, dims=[0, 1, 4, 5, 7], depth=4, width=2):
data = pods.datasets.breastoriginal(dims=dims)
X, Y, XLabel = data['X'], data['Y'], data['XLabel']
d = GPCData(X,
Y,
XLabel=XLabel,
YLabel=['no breast cancer', 'with breast cancer'])
print "\n\nWisconsin Breast Cancer dataset"
print "=====\nData size: D = %d, N = %d." % (d.getDim(), d.getNum())
search = GPCSearch(data=d, max_depth=depth, beam_width=width)
best, best1d = search.search()
constker = search.baseline()
report = GPCReport(name='Wisconsin',
history=best,
best1d=best1d,
constkernel=constker)
report.export()
def cleveland(self, dims=[0, 3, 4, 7, 9, 11], depth=5, width=2):
data = pods.datasets.cleveland(dims=dims)
X, Y, XLabel = data['X'], data['Y'], data['XLabel']
d = GPCData(X,
Y,
XLabel=XLabel,
YLabel=['no heart disease', 'with heart disease'])
print "\n\nCleveland Heart Disease dataset"
print "=====\nData size: D = %d, N = %d." % (d.getDim(), d.getNum())
search = GPCSearch(data=d, max_depth=4, beam_width=2)
best, best1d = search.search()
constker = search.baseline()
report = GPCReport(name='Cleveland',
history=best,
best1d=best1d,
constkernel=constker)
report.export()
# Copyright (c) 2015, Qiurui He
# Department of Engineering, University of Cambridge
import numpy as np
import pods
from gpcdata import GPCData
from gpcsearch import GPCSearch
# 1D
print "\n=====\n1D test:"
X = np.random.uniform(1, 10, (200, 1))
Y = np.zeros((200, 1))
Y[150:, :] = 1
d = GPCData(X, Y)
print d
search = GPCSearch(data=d, max_depth=5, beam_width=2)
best, best1d = search.search()
print "\n=====\nBaseline:"
ck = search.baseline()
print ck
print ck.getGPyKernel()
ck.draw('./imgs/gpcsearchtest', active_dims_only=True)
print "cverror = {}".format(ck.error())
print "error = {}".format(ck.misclassifiedPoints()['X'].shape[0] /
float(ck.data.X.shape[0]))
# 2D
print "\n=====\n2D test:"
data = pods.datasets.crescent_data(seed=496)
X = data['X']
# Copyright (c) 2015, Qiurui He
# Department of Engineering, University of Cambridge
import numpy as np
import pods
from gpcdata import GPCData
from gpcreport import GPCReport
from gpcsearch import GPCSearch
data = pods.datasets.pima()
X = data['X'][:250,[1,5,6,7]]
Y = data['Y'][:250]
d = GPCData(X, Y, XLabel=['plasma glucose', 'BMI', 'pedigree function', 'age'])
print "\n=====\nData size: D = %d, N = %d." % (d.getDim(), d.getNum())
search = GPCSearch(data=d, max_depth=3, beam_width=1)
best, best1d = search.search()
print "\n=====\nBaseline:"
ck = search.baseline()
print ck
print "error = {}".format(ck.error())
report = GPCReport(name='Pima', history=best, best1d=best1d, constkernel=ck)
report.export()
# data = pods.datasets.iris()
# X = data['X']
# Y = data['Y']
