def classify(nnuf, X, Y, train_index, test_index):
svm_X = []
num_candidates = 0
for i in train_index:
nbrs = []
for xi in X[i]:
idx, mag, num_c = nnuf.index(xi, detail=True)
num_candidates += num_c
nbrs.append(idx)
svm_X.append(util.bow(nbrs, D_atoms))
svm = LinearSVC()
svm.fit(svm_X, Y[train_index])
svm_X = []
for i in test_index:
nbrs = []
for xi in X[i]:
idx, mag, num_c = nnuf.index(xi, detail=True)
num_candidates += num_c
nbrs.append(idx)
svm_X.append(util.bow(nbrs, D_atoms))
Y_pred = svm.predict(svm_X)
return accuracy_score(Y[test_index], Y_pred), num_candidates
storage (e.g. 'mini' or 'half')
output_path
chunk_size (size in bytes, -1 for no chunks)
'''
args = json.loads(sys.argv[1])
storage = name_to_storage(args['storage'])
KMeans_tr_size = 200000
X, Y, X_normal = read_dataset(args['tr_folder_path'], args['dtype'])
pipe = pscgen.Pipeline(100, 12)
pipe.fit(X, Y, args['D_atoms'], args['alpha'], args['beta'], storage)
cl1, cl2, cl3 = [], [], []
for i in xrange(len(X)):
x = util.bow(pipe.nnu.index(X[i]), args['D_atoms'])
cl1.append(pipe.svm.predict(x)[0])
cl2.append(pipe.svm.classes_[classify(x, pipe.svm.coef_,
pipe.svm.intercept_, 13)])
cl3.append(pipe.classify(X_normal[i]))
print len([i for i, j in zip(cl1, cl3) if i == j])
assert False
X_Kmeans = np.vstack(X)[:KMeans_tr_size]
# Train D using KMeans
D = KMeans(n_clusters=args['D_atoms'], init_size=args['D_atoms']*3)
D.fit(X_Kmeans)
#nns
print 'Nearest Neighbor'
for i, N in enumerate(Ns):
D = KMeans(n_clusters=N, init_size=N*3)
D.fit(X_tr_Kmeans)
D = D.cluster_centers_
D = util.normalize(D)
D_mean = np.mean(D, axis=0)
D = D - D_mean
svm_nns_xs_tr, svm_nns_xs_t = [], []
for x in X_tr:
x = util.normalize(x)
x = x - D_mean
nbrs = np.argmax(np.abs(np.dot(D, x.T)), axis=0)
svm_nns_xs_tr.append(util.bow(nbrs, N))
for x in X_t:
x = util.normalize(x)
x = x - D_mean
nbrs = np.argmax(np.abs(np.dot(D, x.T)), axis=0)
svm_nns_xs_t.append(util.bow(nbrs, N))
acc = util.predict_chi2(svm_nns_xs_tr, Y_tr, svm_nns_xs_t, Y_t)
print N, acc
nns_dists.append(acc)
D = KMeans(n_clusters=NNU_N, init_size=NNU_N*3)
D.fit(X_tr_Kmeans)
D = D.cluster_centers_
