def try_train_HMM_on_Data():
N = 8
M = OBSERVE_STATE_LIST.__len__()
a_matrix = init_A(N)
b_matrix = init_B(N, M)
pi = init_PI(N)
# print(a_matrix);
# print(b_matrix);
# print(pi);
_data = getHMMData()
o_sequence_List = []
for index in range(_data.__len__()):
o_sequence_List.append(_data[index][1])
a_matrix, b_matrix, pi = baum_welch_multipleObservation(
a_matrix, b_matrix, pi, o_sequence_List, None, 10)
# print(normalize_matrix_line(shelterSmallValue(a_matrix,1e-5,0)));
# print(normalize_matrix_line(shelterSmallValue(b_matrix,1e-5,0)));
# print(normalize_matrix_line(shelterSmallValue(np.array([pi]),1e-5,0))[0]);
print(compute_bic_of_HMM(a_matrix, b_matrix, pi, o_sequence_List))
def convert_index_to_Uid():
_data = getHMMData()
index_to_uid_dic = {}
for index in range(_data.__len__()):
index_to_uid_dic[index] = _data[index][0]
_out_data = copy.deepcopy(cluster_result)
for cluster_index in range(cluster_result.__len__()):
for item_index, value in enumerate(cluster_result[cluster_index]):
_out_data[cluster_index][item_index] = index_to_uid_dic[value]
print(_out_data)
for i in range(K - 1):
for j in range(1, K):
is_pre_same_cluster = pre_cluster_result_dic[
i] == pre_cluster_result_dic[j]
is_new_same_cluster = new_cluster_result_dic[
i] == new_cluster_result_dic[j]
disMatchCount += is_new_same_cluster ^ is_pre_same_cluster
return disMatchCount > 0
if __name__ == "__main__":
# 数据准备
M = OBSERVE_STATE_LIST.__len__()
_data = getHMMData()
o_sequence_List = []
for index in range(_data.__len__()):
o_sequence_List.append(_data[index][1])
a_matrix_list, b_matrix_list, pi_list, data_of_clusters = do_HMM_Cluster_On_data_with_dN(
o_sequence_List, 3, M, show_progress=True)
#a_matrix_list, b_matrix_list, pi_list, data_of_clusters = do_HMM_Cluster_On_data(o_sequence_List,3,4,M,show_progress=True)
print(a_matrix_list)
print(b_matrix_list)
print(pi_list)
print(data_of_clusters)
PMI = compute_PMI_of_HMM_Clusters(a_matrix_list, b_matrix_list, pi_list,
data_of_clusters, o_sequence_List)