def knn_learn(nneighbors=1,
data_train=np.array([]),
target_train=np.array([]),
data_test=np.array([])):
"""
estimate position using the K neareast neighbors (KNN) technique
Parameters
----------
nneighbors : int
default = 1
data_train : numpy.ndarray
default = array([])
target_train : numpy.ndarray
default = array([])
data_test : numpy.ndarray
default = array([])
Returns
-------
targets : numpy.ndarray
"""
clf = ngb.NeighborsClassifier(nneighbors).fit(data_train, target_train)
targets = clf.predict(data_test)
return targets
def test_pipeline():
# check that LocallyLinearEmbedding works fine as a Pipeline
from sklearn import pipeline, datasets
iris = datasets.load_iris()
clf = pipeline.Pipeline([('filter', manifold.LocallyLinearEmbedding()),
('clf', neighbors.NeighborsClassifier())])
clf.fit(iris.data, iris.target)
assert_lower(.7, clf.score(iris.data, iris.target))
def test_pipeline():
# check that Isomap works fine as a transformer in a Pipeline
iris = datasets.load_iris()
clf = pipeline.Pipeline([('isomap', manifold.Isomap()),
('neighbors_clf', neighbors.NeighborsClassifier())
])
clf.fit(iris.data, iris.target)
assert_lower(.7, clf.score(iris.data, iris.target))
def bench_skl(X, y, T, valid):
#
# .. scikits.learn ..
#
from sklearn import neighbors
start = datetime.now()
clf = neighbors.NeighborsClassifier(n_neighbors=n_neighbors,
algorithm='brute_inplace')
clf.fit(X, y)
score = np.mean(clf.predict(T) == valid)
return score, datetime.now() - start
def run():
# pdb.set_trace();
numComponents = 10;
numCodeWords = 200;
#data loading and normalization
print 'loading data'
pcaMean = loadtxt('pcamean.txt');
pcaComps = loadtxt('pcacomps.txt');
mean = loadtxt('mean.txt');
std = loadtxt('std.txt');
data = loadtxt('/home/mohsen/Downloads/shotFiles/shot.txt');
# mean = data.mean(axis=0);
# std = sqrt(data.var(axis=0));
# std [where(std==0)] = 1;
data = (data - mean ) / std;
#dimension reduction
print 'calculating pca'
pca = decomposition.PCA(n_components=numComponents);
pca.fit(data);
projection = pca.transform(data);
# savetxt('mean.txt',mean);
# savetxt('std.txt',std);
# savetxt('pcamean.txt',pca.mean_);
# savetxt('pcacomps.txt',pca.components_);
# raw_input(':');
#codebook generation
print 'fnding k-means'
# kmeans = cluster.KMeans(k=numCodeWords);
# kmeans.fit(projection);
# codeBook = kmeans.cluster_centers_;
# savetxt('codebook.txt',codeBook)
codeBook = loadtxt('codebook.txt');
#building histogram for eahc object
print 'building histograms'
root = '/home/mohsen/Downloads/shotFiles/';
f = open(root + 'listSingleSample.txt','r');
data = zeros((1,shape(codeBook)[0]));
target = [0];
for line in f:
shotFileName = root + line.replace('\n','');
objectFeatures = loadtxt(shotFileName);
objectLabel = labelFromFileName(shotFileName);
if objectLabel==-1:
print 'Error: Cannot determine label from file name!!!'
return;
objectFeatures = objectFeatures - mean;
objectFeatures = objectFeatures / std;
#objectPca = pca.transform(objectFeatures);
temp = objectFeatures - pcaMean;
objectPca = dot(temp, pcaComps.T);
objectHist = zeros((1,shape(codeBook)[0]));
for idx in range(shape(objectPca)[0]):
dist = sum( (codeBook - objectPca[idx,:])**2 , axis=1 );
objectHist[0,argmin(dist)] = objectHist[0,argmin(dist)] + 1;
objectHist = objectHist / sum(objectHist);
data = concatenate((data,objectHist),axis=0);
target.append(objectLabel);
# pdb.set_trace();
# dataTrain = concatenate((data,target),axis=1);
# savetxt('dataTrain.txt',dataTrain);
train = concatenate( (data[1::3,:],data[2::3,:]),axis=0 );
test = data[3::3,:];
traint = concatenate((target[1::3],target[2::3]));
testt = target[3::3];
print 'knn'
knn1 = neighbors.NeighborsClassifier(1)
knn1.fit(train, traint)
out = knn1.predict(test);
correct = sum(out==testt)/float(shape(out)[0])
print "correct percent is:", correct
printConfMat(testt,out);