def __test__():
dataSet, labels = kNN.createDataSet()
inX = array([1.2, 1.1])
k = 3
outputLabel = kNN.kNNClassify(inX, dataSet, labels, 3)
print("Your input is:", inX, "and classified to class: ", outputLabel)
inX = array([0.1, 0.3])
outputLabel = kNN.kNNClassify(inX, dataSet, labels, 3)
print("Your input is:", inX, "and classified to class: ", outputLabel)
def kNNtest():
# 生成数据集和类别标签
dataSet, labels = kNN.createDataSet()
# 定义一个未知类别的数据
with open('test_24.json', 'r') as file:
for line in file:
line = json.loads(line)
testid = line[0]
testtarget = line[1:]
testX = array([testtarget])
# 调用分类函数对未知数据分类
outputLabel = kNN.kNNClassify(testX, dataSet, labels, 1)
result = [outputLabel, testid]
with open('testPredict24.json', 'a') as file:
file.write(json.dumps(result) + '\n')
def compute_accuracy(test_proto_visual, test_visual, test_proto2label,
test_x2label):
# test_proto: [50, 312]
# test visual: [2993, 1024]
# test_proto2label: proto2label [50]
# test_x2label: x2label [2993]
outpred = [0] * test_visual.shape[0]
for i in range(test_visual.shape[0]):
outputLabel = kNN.kNNClassify(test_visual[i, :],
test_proto_visual.cpu().data.numpy(),
test_proto2label, 1)
outpred[i] = outputLabel
outpred = np.array(outpred)
acc = np.equal(outpred, test_x2label).mean()
return acc
def compute_accuracy(test_att, test_visual, test_id, test_label):
global left_a2
att_pre = sess.run(left_a2, feed_dict={att_features: test_att})
test_id = np.squeeze(np.asarray(test_id))
outpre = [0] * 2933
test_label = np.squeeze(np.asarray(test_label))
test_label = test_label.astype("float32")
for i in range(2933):
outputLabel = kNN.kNNClassify(test_visual[i, :], att_pre, test_id, 1)
outpre[i] = outputLabel
correct_prediction = tf.equal(outpre, test_label)
accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))
result = sess.run(accuracy,
feed_dict={
att_features: test_att,
visual_features: test_visual
})
return result
def compute_accuracy(test_att, test_visual, test_id, test_label):
global left_a2
att_pre = sess.run(left_a2, feed_dict={att_features: test_att})
test_id = np.squeeze(np.asarray(test_id))
outpre = [0] * test_visual.shape[0] # CUB 2933
test_label = np.squeeze(np.asarray(test_label))
test_label = test_label.astype("float32")
for i in range(test_visual.shape[0]): # CUB 2933
outputLabel = kNN.kNNClassify(test_visual[i, :], att_pre, test_id, 1)
outpre[i] = outputLabel
# compute averaged per class accuracy
outpre = np.array(outpre, dtype='int')
unique_labels = np.unique(test_label)
acc = 0
for l in unique_labels:
idx = np.nonzero(test_label == l)[0]
acc += accuracy_score(test_label[idx], outpre[idx])
acc = acc / unique_labels.shape[0]
return acc
def compute_accuracy(test_att, test_visual, test_id, test_label):
# test_att: [2993, 312]
# test viaual: [2993, 1024]
# test_id: att2label [50]
# test_label: x2label [2993]
test_att = Variable(torch.from_numpy(test_att).float().cuda())
att_pred = forward(test_att)
outpred = [0] * 2933
test_label = test_label.astype("float32")
# att_pre [50, 1024],
# test_visual: [2993, 1024]
# test_id : [50]
for i in range(2933):
outputLabel = kNN.kNNClassify(test_visual[i, :],
att_pred.cpu().data.numpy(), test_id, 1)
outpred[i] = outputLabel
outpred = np.array(outpred)
acc = np.equal(outpred, test_label).mean()
return acc
import kNN
from numpy import *
dataSet, labels = kNN.createDataSet()
testX = array([1.9, 3.2])
k = 3
outputLabel = kNN.kNNClassify(testX, dataSet, labels, 3)
print("Your input is:", testX, "and classified to class: ", outputLabel)
testX = array([4.1, 3.3])
outputLabel = kNN.kNNClassify(testX, dataSet, labels, 3)
print("Your input is:", testX, "and classified to class: ", outputLabel)
# -*- coding: utf-8 -*- # writer : lgy # data : 2017-08-19 import kNN from numpy import * dataSet, labels = kNN.createDataSet() testX = array([1.2, 1.0]) k = 3 outputLabel = kNN.kNNClassify(testX, dataSet, labels, 3) print "Your input is:", testX, "and classified to class: ", outputLabel testX = array([0.1, 0.3]) outputLabel = kNN.kNNClassify(testX, dataSet, labels, 3) print "Your input is:", testX, "and classified to class: ", outputLabel
def compute_accuracy(generator, test_semantic, test_visual, test_classIdx, \
test_videoLabel, use_z, n_gene_perC, classifier_type, \
n_epoch_sftcls):
'''Evaluation'''
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
n_generation_perClass = n_gene_perC # how many samples are generated by G per class
nb_samples = test_visual.shape[0]
test_semantic = torch.from_numpy(test_semantic).float().to(device)
if use_z == 'true':
z = torch.randn(test_semantic.size()[0], z_dim).to(device)
syn_vi_fea = generator(test_semantic, z)
else:
syn_vi_fea = generator(test_semantic)
## kNN as a classifier during testing
if classifier_type == 'knn':
outpred = [0] * nb_samples
test_videoLabel = test_videoLabel.astype("float32")
for i in range(nb_samples):
outputLabel = kNN.kNNClassify(test_visual[i, :],
syn_vi_fea.cpu().data.numpy(),
test_classIdx, 5)
outpred[i] = outputLabel
outpred = np.array(outpred)
acc = np.equal(outpred, test_videoLabel).mean()
## kernel svm as a classifier during testing
elif classifier_type == 'svm' or 'softmax':
expend_testvideoLabel = test_classIdx
for cnt in range(n_generation_perClass - 1):
if use_z == 'true':
z = torch.randn(test_semantic.size()[0], z_dim).to(device)
temp_syn_vi_fea = generator(test_semantic, z)
else:
temp_syn_vi_fea = generator(test_semantic)
syn_vi_fea = torch.cat((syn_vi_fea, temp_syn_vi_fea), dim=0)
expend_testvideoLabel = np.hstack(
(expend_testvideoLabel, test_classIdx))
assert syn_vi_fea.size(0) == expend_testvideoLabel.shape[0]
syn_vi_fea = syn_vi_fea.cpu().detach().numpy()
dataset_obj = dict()
dataset_obj['tr_data'] = syn_vi_fea
dataset_obj['tr_label'] = expend_testvideoLabel
dataset_obj['te_data'] = test_visual
dataset_obj['te_label'] = test_videoLabel
if classifier_type == 'svm':
ksvm = MyKSVM(dataset_obj)
acc1 = ksvm.linear_kernel()
acc2 = ksvm.nonlinear_kernel(
) # if the degree equals to 3, may raise a MemoryError
acc3 = ksvm.poly_kernel()
acc4 = ksvm.RBF_kernel()
acc = max((acc1, acc2, acc3, acc4))
elif classifier_type == 'softmax':
n_class = len(test_classIdx)
n_epoch = n_epoch_sftcls
batch_size = -1
lr = 1e-3
sft_cls = fit_softmaxClassifier(n_class, dataset_obj, n_epoch,
batch_size, lr)
te_y = dataset_obj['te_label']
global_label = set(te_y)
globalLabel2localLabel = dict()
for global_idx, local_idx in zip(global_label,
list(range(len(global_label)))):
globalLabel2localLabel[global_idx] = local_idx
te_y = np.array([globalLabel2localLabel[i] for i in te_y])
te_x = torch.from_numpy(dataset_obj['te_data']).float()
te_y = torch.from_numpy(te_y).long()
sft_cls.cpu()
te_y_hat = sft_cls(te_x)
acc = (np.argmax(te_y_hat.detach().numpy(), axis=1)
== te_y.detach().numpy()).sum() / float(te_y.size()[0])
return acc
#!/usr/bin/env python
#_*_coding:utf-8_*_
#kNN算法，调用示例。数据集在loadDataSet中设定了，也可以从文件读取
from numpy import mat
import kNN
if __name__ == "__main__":
print("kNN近邻程序,输入：A B，如 1 2")
data = mat(input()).tolist()
trainData, yLabel = kNN.loadDataSet()
result = kNN.kNNClassify(data, trainData, yLabel, 2, k=3)
print(result)
