def drawGraphForDatasets(datasets, fileName, item, trainBlockSizes, metric, ylim = []):
plt.subplot(111)
plt.figure(item)
plt.title(metric)
plt.xlabel("% of dataset")
plt.ylabel("score: %s" % metric)
plt.grid()
for load in datasets:
# load it lazy
data, label, desc = load()
# for test - make dataset smaller
initialReduceBlockSize = np.arange(0.5, 0.7, 0.1)
trainDataBlocks, trainLabelBlocks, testDataBlocks, testLabelBlocks = factory.splitDatasetInBlocks(data, np.array(label), initialReduceBlockSize, testSetPercentage)
data = trainDataBlocks[0][0]
label = trainLabelBlocks[0][0]
analyze(data, label)
maxItemsInDataset = len(label)
testSetPercentage = 0.02
trainDataBlocks, trainLabelBlocks, testDataBlocks, testLabelBlocks = factory.splitDatasetInBlocks(data, np.array(label), trainBlockSizes, testSetPercentage)
x = list()
y = list()
for i in range(len(trainDataBlocks)):
trainData = trainDataBlocks[i]
trainLabel = trainLabelBlocks[i]
# testData = testDataBlocks[i]
# testLabel = testLabelBlocks[i]
numInstances = np.shape(trainData[0])
score = calcScore(metric, trainData[0], trainLabel[0])
xPercentage = (numInstances[0] * 100) / maxItemsInDataset
x.append(xPercentage)
#y.append(float("%.4f" % score))
y.append(score)
#print "x:%s, y:%s" % (numInstances[0], score)
print "------------------------"
print y
print np.mean(y)
print "------------------------"
plt.plot(x, y, label=desc)
plt.legend(loc="best")
if len(ylim) > 0:
plt.ylim(ylim)
plt.savefig("performance/output/%s_%s.png" % (fileName, metric), dpi=320)
def loadData(experiment):
if experiment.has_key("size"):
size = experiment["size"]
else:
size = 0
data, label, description, reduce = experiment["dataset"]()
if size > 0:
initialReduceBlockSize = np.arange(size, size + 0.2, 0.1)
testSetPercentage = 0.2
trainDataBlocks, trainLabelBlocks, testDataBlocks, testLabelBlocks = data_factory.splitDatasetInBlocks(
data, np.array(label), initialReduceBlockSize, testSetPercentage)
data = trainDataBlocks[0][0]
label = trainLabelBlocks[0][0]
# if required (cancer datasets) perform binary encoding
if experiment['binary_encode']:
print "perform binary encode"
analyze(data, label, "before encode")
# encode features (one-hot-encoder / dummy coding)
enc = OneHotEncoder()
enc.fit(data)
data = enc.transform(data).toarray()
analyze(data, label, "after encode")
return data, label, description, reduce
def loadData(experiment):
if experiment.has_key("size"):
size = experiment["size"]
else:
size = 0
data, label, description, reduce = experiment["dataset"]()
if size > 0:
initialReduceBlockSize = np.arange(size, size+0.2, 0.1)
testSetPercentage = 0.2
trainDataBlocks, trainLabelBlocks, testDataBlocks, testLabelBlocks = data_factory.splitDatasetInBlocks(data, np.array(label), initialReduceBlockSize, testSetPercentage)
data = trainDataBlocks[0][0]
label = trainLabelBlocks[0][0]
# if required (cancer datasets) perform binary encoding
if experiment['binary_encode']:
print "perform binary encode"
analyze(data, label, "before encode")
# encode features (one-hot-encoder / dummy coding)
enc = OneHotEncoder()
enc.fit(data)
data = enc.transform(data).toarray()
analyze(data, label, "after encode")
return data, label, description, reduce
def getLearningCurve(data, label):
trainDataBlocks, trainLabelBlocks, testDataBlocks, testLabelBlocks = data_factory.splitDatasetInBlocks(data, np.array(label), trainBlockSizes, testSetPercentage)
x = list()
yTrain = list()
yTest = list()
for i in range(len(trainDataBlocks)):
trainData = trainDataBlocks[i]
trainLabel = trainLabelBlocks[i]
testData = testDataBlocks[i]
testLabel = testLabelBlocks[i]
trainScores = list()
testScores = list()
for j in range(len(trainData)):
trainS, testS = rocAuc(trainData[j], trainLabel[j], testData[j], testLabel[j])
trainScores.append(trainS)
testScores.append(testS)
yTrain.append(np.mean(trainScores))
yTest.append(np.mean(testScores))
if hasattr(trainData[0], "indices"):
numInstances = np.shape(trainData[0])
numInstances = numInstances[0]
else:
numInstances = np.shape(trainData)
numInstances = numInstances[1]
x.append(numInstances)
return x, yTrain, yTest
def getXYForDataLabel(data, label):
maxItemsInDataset = len(label)
trainDataBlocks, trainLabelBlocks, testDataBlocks, testLabelBlocks = data_factory.splitDatasetInBlocks(data, np.array(label), trainBlockSizes, testSetPercentage)
x = list()
y = list()
for i in range(len(trainDataBlocks)):
trainData = trainDataBlocks[i]
trainLabel = trainLabelBlocks[i]
scores = list()
for j in range(len(trainData)):
s = rocAuc(trainData[j], trainLabel[j])
scores.append(s)
numInstances = np.shape(trainData[0])
xPercentage = (numInstances[0] * 100) / maxItemsInDataset
x.append(xPercentage)
y.append(np.mean(scores))
return x, y
estimatedDimension = mle(data)
print "estimated dimension without binary encoding: %s" % estimatedDimension
enc = OneHotEncoder()
enc.fit(data)
data = enc.transform(data).toarray()
analyze(data, label, "after binary encode")
estimatedDimensionWithBinaryEncode = mle(data)
print "estimated dimension with binary encoding: %s" % estimatedDimensionWithBinaryEncode
print
'''
data, label, desc, size = df.loadFirstPlistaDataset()
trainBlockSizes = np.arange(0.001, 0.005, 0.001)
testSetPercentage = 0.2
trainDataBlocks, trainLabelBlocks, testDataBlocks, testLabelBlocks = df.splitDatasetInBlocks(
data, np.array(label), trainBlockSizes, testSetPercentage)
# estimate for different sizes of the data set
x = list()
yDuration = list()
yDimension = list()
for i in range(len(trainDataBlocks)):
estimatedDimension = list()
duration = list()
shape = np.shape(trainDataBlocks[i][0])
x.append(shape[0])
for j in range(len(trainDataBlocks[i])):
print j
data = trainDataBlocks[i][j]
def getThirdPlistaData():
data, label, _, _ = df.loadThridPlistaDataset()
initialReduceBlockSize = np.arange(0.01, 0.3, 0.1)
trainDataBlocks, trainLabelBlocks, testDataBlocks, testLabelBlocks = df.splitDatasetInBlocks(
data, np.array(label), initialReduceBlockSize, 0.1)
return trainDataBlocks[0][0], trainLabelBlocks[0][0], _, _
import numpy as np
import time
from sklearn import linear_model
import data_factory as df
import matplotlib.pyplot as plt
import dr
plt.figure()
plt.xlabel('dimensions')
plt.ylabel('duration (seconds)')
plt.title('ROC Curves')
data, label, desc, size = df.loadFirstPlistaDataset()
initialReduceBlockSize = np.arange(0.1, 0.7, 0.1)
trainDataBlocks, trainLabelBlocks, testDataBlocks, testLabelBlocks = df.splitDatasetInBlocks(data, np.array(label), initialReduceBlockSize, 0.1)
data = trainDataBlocks[0][0]
label = trainLabelBlocks[0][0]
x = list()
sparse_y = list()
dense_y = list()
algos = dr.getAllAlgosExlude(["tsne"])
algodurations = {}
dimension_range = np.arange(50, 250, 50)
for i in dimension_range: # dimensions
x.append(i)
def getThirdPlistaData():
data, label, _, _ = df.loadThridPlistaDataset()
initialReduceBlockSize = np.arange(0.01, 0.3, 0.1)
trainDataBlocks, trainLabelBlocks, testDataBlocks, testLabelBlocks = df.splitDatasetInBlocks(data, np.array(label), initialReduceBlockSize, 0.1)
return trainDataBlocks[0][0], trainLabelBlocks[0][0],_,_
numInstances = np.shape(trainData[0])
xPercentage = (numInstances[0] * 100) / maxItemsInDataset
x.append(xPercentage)
y.append(np.mean(scores))
return x, y
dataSets = data_factory.getSmallDatasets()
for i in range(len(dataSets)):
load = dataSets[i]
data, label, desc, size = load()
if size > 0:
initialReduceBlockSize = np.arange(size, size+0.2, 0.1)
trainDataBlocks, trainLabelBlocks, testDataBlocks, testLabelBlocks = data_factory.splitDatasetInBlocks(data, np.array(label), initialReduceBlockSize, testSetPercentage)
data = trainDataBlocks[0][0]
label = trainLabelBlocks[0][0]
print np.shape(data)
print np.shape(label)
analyze(data, label)
plt.subplot(111)
plt.figure(i)
plt.title("%s - %s" % (desc, metric))
plt.xlabel("% of dataset")
plt.ylabel(metric)
plt.grid()
