def test_10cv():
"""
"""
file_name = "D:/Data/OneDrive/文档/Code/MIL1/Data/Text/alt_atheism.mat"
"""======================================================="""
loops = 5
# tr_f1_k, tr_acc_k, tr_roc_k = np.zeros(loops), np.zeros(loops), np.zeros(loops)
# tr_f1_s, tr_acc_s, tr_roc_s = np.zeros(loops), np.zeros(loops), np.zeros(loops)
# tr_f1_j, tr_acc_j, tr_roc_j = np.zeros(loops), np.zeros(loops), np.zeros(loops)
te_f1_k, te_acc_k, te_roc_k = np.zeros(loops), np.zeros(loops), np.zeros(loops)
te_f1_s, te_acc_s, te_roc_s = np.zeros(loops), np.zeros(loops), np.zeros(loops)
te_f1_j, te_acc_j, te_roc_j = np.zeros(loops), np.zeros(loops), np.zeros(loops)
print("=================================================")
for i in range(loops):
mil = MILDM(file_name)
classifier = Classify(["knn", "svm", "j48"], ["f1_score", "acc", "roc"])
data_iter = mil.get_discriminative_ins()
te_per = classifier.test(data_iter)
# tr_f1_k[i], tr_acc_k[i], tr_roc_k[i] = tr_per["knn"][0], tr_per["knn"][1], tr_per["knn"][2]
# tr_f1_s[i], tr_acc_s[i], tr_roc_s[i] = tr_per["svm"][0], tr_per["svm"][1], tr_per["svm"][2]
# tr_f1_j[i], tr_acc_j[i], tr_roc_j[i] = tr_per["j48"][0], tr_per["j48"][1], tr_per["j48"][2]
te_f1_k[i], te_acc_k[i], te_roc_k[i] = te_per["knn"][0], te_per["knn"][1], te_per["knn"][2]
te_f1_s[i], te_acc_s[i], te_roc_s[i] = te_per["svm"][0], te_per["svm"][1], te_per["svm"][2]
te_f1_j[i], te_acc_j[i], te_roc_j[i] = te_per["j48"][0], te_per["j48"][1], te_per["j48"][2]
print("%.4lf, %.4lf, %.4lf; %.4lf, %.4lf, %.4lf; %.4lf, %.4lf, %.4lf; \n"
% (te_f1_k[i], te_acc_k[i], te_roc_k[i],
te_f1_s[i], te_acc_s[i], te_roc_s[i],
te_f1_j[i], te_acc_j[i], te_roc_j[i]
), end=" ")
# print("%.4lf %.4lf %.4lf %.4lf %.4lf %.4lf "
# "%.4lf %.4lf %.4lf %.4lf %.4lf %.4lf "
# "%.4lf %.4lf %.4lf %.4lf %.4lf %.4lf " % (np.sum(tr_f1_k) / loops, np.std(tr_f1_k),
# np.sum(tr_acc_k) / loops, np.std(tr_acc_k),
# np.sum(tr_roc_k) / loops, np.std(tr_roc_k),
# np.sum(tr_f1_s) / loops, np.std(tr_f1_s),
# np.sum(tr_acc_s) / loops, np.std(tr_acc_s),
# np.sum(tr_roc_s) / loops, np.std(tr_roc_s),
# np.sum(tr_f1_j) / loops, np.std(tr_f1_j),
# np.sum(tr_acc_j) / loops, np.std(tr_acc_j),
# np.sum(tr_roc_j) / loops, np.std(tr_roc_j)), end="")
print("knn-f1 std knn-acc std knn-roc std svm-f1 std svm-acc std svm-roc std "
"j48-f1 std j48-acc std j48-roc std")
print("%.4lf %.4lf %.4lf %.4lf %.4lf %.4lf "
"%.4lf %.4lf %.4lf %.4lf %.4lf %.4lf "
"%.4lf %.4lf %.4lf %.4lf %.4lf %.4lf " % (np.sum(te_f1_k) / loops, np.std(te_f1_k),
np.sum(te_acc_k) / loops, np.std(te_acc_k),
np.sum(te_roc_k) / loops, np.std(te_roc_k),
np.sum(te_f1_s) / loops, np.std(te_f1_s),
np.sum(te_acc_s) / loops, np.std(te_acc_s),
np.sum(te_roc_s) / loops, np.std(te_roc_s),
np.sum(te_f1_j) / loops, np.std(te_f1_j),
np.sum(te_acc_j) / loops, np.std(te_acc_j),
np.sum(te_roc_j) / loops, np.std(te_roc_j)), end="")
def test_10cv():
"""
"""
po_label = 9
file_name = "mnist" + str(
po_label
) + ".none" # "D:/Data/OneDrive/文档/Code/MIL1/Data/Text/talk_religion_misc.mat"
mnist_path = "D:/Data/OneDrive/文档/Code/MIL1/Data"
"""======================================================="""
loops = 5
# tr_f1_k, tr_acc_k, tr_roc_k = np.zeros(loops), np.zeros(loops), np.zeros(loops)
# tr_f1_s, tr_acc_s, tr_roc_s = np.zeros(loops), np.zeros(loops), np.zeros(loops)
# tr_f1_j, tr_acc_j, tr_roc_j = np.zeros(loops), np.zeros(loops), np.zeros(loops)
te_f1_k, te_acc_k, te_roc_k = np.zeros(loops), np.zeros(loops), np.zeros(
loops)
te_f1_s, te_acc_s, te_roc_s = np.zeros(loops), np.zeros(loops), np.zeros(
loops)
te_f1_j, te_acc_j, te_roc_j = np.zeros(loops), np.zeros(loops), np.zeros(
loops)
print("=================================================")
print("BAMIC with %s" % file_name.split("/")[-1].split(".")[0])
bag_space = MnistLoader(seed=1, po_label=po_label,
mnist_path=mnist_path).bag_space
mil = BaMic(file_name, bag_space=bag_space)
for i in range(loops):
classifier = Classify(["knn", "svm", "j48"],
["f1_score", "acc", "roc"])
data_iter = mil.get_mapping()
te_per = classifier.test(data_iter)
# tr_f1_k[i], tr_acc_k[i], tr_roc_k[i] = tr_per["knn"][0], tr_per["knn"][1], tr_per["knn"][2]
# tr_f1_s[i], tr_acc_s[i], tr_roc_s[i] = tr_per["svm"][0], tr_per["svm"][1], tr_per["svm"][2]
# tr_f1_j[i], tr_acc_j[i], tr_roc_j[i] = tr_per["j48"][0], tr_per["j48"][1], tr_per["j48"][2]
te_f1_k[i], te_acc_k[i], te_roc_k[i] = te_per["knn"][0], te_per["knn"][
1], te_per["knn"][2]
te_f1_s[i], te_acc_s[i], te_roc_s[i] = te_per["svm"][0], te_per["svm"][
1], te_per["svm"][2]
te_f1_j[i], te_acc_j[i], te_roc_j[i] = te_per["j48"][0], te_per["j48"][
1], te_per["j48"][2]
print(
"%.4lf, %.4lf, %.4lf; %.4lf, %.4lf, %.4lf; %.4lf, %.4lf, %.4lf; \n"
% (te_f1_k[i], te_acc_k[i], te_roc_k[i], te_f1_s[i], te_acc_s[i],
te_roc_s[i], te_f1_j[i], te_acc_j[i], te_roc_j[i]),
end=" ")
# print("%.4lf %.4lf %.4lf %.4lf %.4lf %.4lf "
# "%.4lf %.4lf %.4lf %.4lf %.4lf %.4lf "
# "%.4lf %.4lf %.4lf %.4lf %.4lf %.4lf " % (np.sum(tr_f1_k) / loops, np.std(tr_f1_k),
# np.sum(tr_acc_k) / loops, np.std(tr_acc_k),
# np.sum(tr_roc_k) / loops, np.std(tr_roc_k),
# np.sum(tr_f1_s) / loops, np.std(tr_f1_s),
# np.sum(tr_acc_s) / loops, np.std(tr_acc_s),
# np.sum(tr_roc_s) / loops, np.std(tr_roc_s),
# np.sum(tr_f1_j) / loops, np.std(tr_f1_j),
# np.sum(tr_acc_j) / loops, np.std(tr_acc_j),
# np.sum(tr_roc_j) / loops, np.std(tr_roc_j)), end="")
print(
"knn-f1 std knn-acc std knn-roc std svm-f1 std svm-acc std svm-roc std "
"j48-f1 std j48-acc std j48-roc std")
print("%.4lf %.4lf %.4lf %.4lf %.4lf %.4lf "
"%.4lf %.4lf %.4lf %.4lf %.4lf %.4lf "
"%.4lf %.4lf %.4lf %.4lf %.4lf %.4lf " %
(np.sum(te_f1_k) / loops, np.std(te_f1_k), np.sum(te_acc_k) / loops,
np.std(te_acc_k), np.sum(te_roc_k) / loops, np.std(te_roc_k),
np.sum(te_f1_s) / loops, np.std(te_f1_s), np.sum(te_acc_s) / loops,
np.std(te_acc_s), np.sum(te_roc_s) / loops, np.std(te_roc_s),
np.sum(te_f1_j) / loops, np.std(te_f1_j), np.sum(te_acc_j) / loops,
np.std(te_acc_j), np.sum(te_roc_j) / loops, np.std(te_roc_j)),
end="")