def generate_features_with_ref_segments(self,
selected_tuple,
patient_no,
subs_flag=False):
#print("Inside train")
if subs_flag: print("Using Substitute")
selected_label = selected_tuple[0]
selected_segment = selected_tuple[1]
s1 = np.array(selected_segment)
F_avg = []
for ref_segment in self.ref_segments[self.ref_label]:
s2 = np.array(ref_segment)
try:
F = feature_gen(s1, s2)
F_avg.append(F)
except Warning:
print("Warning encountered..")
print("*************************")
substitute_tuple = self.extract_random_segments_for_given_patient_during_warning(
selected_label, patient_no)
self.generate_features_with_ref_segments(
substitute_tuple, patient_no, subs_flag=True
) #recursively call this function till Warningless segment is found (in practice Warning is rarely encountered, hence more than 1 recursive call is extremely rare)
return #important, else the local s1(which is the source of Warning) will continue executing, thus calling the functions in except block again and again
self.trainset_list.append((selected_label, np.mean(F_avg, axis=0)))
np.save(f"clf{self.ref_label}.npy", self.trainset_list)
def generate_features_with_ref_segments(self, selected_tuple,patient_no):
selected_label = selected_tuple[0]
selected_segment = selected_tuple[1]
t, s = np.arange(len(selected_segment)), np.array(selected_segment)
#l = 1 if selected_label == self.ref_label else 0
for ref in range(NUM_SLEEP_STAGES):
F_avg = []
for ref_segment in self.ref_segments[ref]:
t1, s1 = t, s
t2, s2 = np.arange(len(ref_segment)), np.array(ref_segment)
# print(s1.shape)
# print(s2.shape)
#converting segments to equal lengths
S1 = s1[np.argwhere((t1 >= min(t2)) & (t1 <= max(t2))).flatten()]
S2 = s2[np.argwhere((t2 >= min(t1)) & (t2 <= max(t1))).flatten()]
# print(s1.shape)
# print(s2.shape)
# print("*************************")
try:
F = feature_gen(t1, S1, t2, S2, ref)
F_avg.append(F)
except Warning:
global no_of_errors_encountered
no_of_errors_encountered+=1
substitute=extract_random_segments_for_given_patient_during_warning(selected_label,patient_no)
self.generate_features_with_ref_segments(substitute,patient_no)
self.testset_dict[ref].append((selected_label, np.mean(F_avg, axis=0)))
def generate_features_with_ref_segments(self, selected_tuple, patient_no):
selected_label = selected_tuple[0]
selected_segment = selected_tuple[1]
t, s = np.arange(len(selected_segment)), np.array(selected_segment)
#l = 1 if selected_label == self.ref_label else 0
F_avg = []
for ref_segment in self.ref_segments[self.ref_label]:
t1, s1 = t, s
t2, s2 = np.arange(len(ref_segment)), np.array(ref_segment)
# print(s1.shape)
# print(s2.shape)
#converting segments to equal lengths
S1 = s1[np.argwhere((t1 >= min(t2)) & (t1 <= max(t2))).flatten()]
S2 = s2[np.argwhere((t2 >= min(t1)) & (t2 <= max(t1))).flatten()]
# print(s1.shape)
# print(s2.shape)
# print("*************************")
try:
F = feature_gen(t1, S1, t2, S2, self.ref_label)
for i in range(len(F)):
if math.isnan(F[i]):
F[i] = 0
F_avg.append(F)
#for i in range(len(F)):
#print(f"{F[i]}({i})", end=", ")
except Warning:
global no_of_errors_encountered
no_of_errors_encountered += 1
substitute = self.extract_random_segments_for_given_patient_during_warning(
selected_label, patient_no)
return self.generate_features_with_ref_segments(
substitute, patient_no)
#print(f"Feature vector:{np.mean(F_avg,axis=0)}")
return selected_label
self.data_list.append((selected_label, np.mean(F_avg, axis=0)))
def generate_features_with_ref_segments(self, selected_tuple, patient_no,
mean, std):
selected_label = selected_tuple[0]
selected_segment = selected_tuple[1]
s1 = np.array(selected_segment)
for ref in range(NUM_SLEEP_STAGES): #extra loop over all ref labels
F_avg = []
for ref_segment in self.ref_segments[ref]:
s2 = np.array(ref_segment)
try:
F = feature_gen(s1, s2, mean, std)
F_avg.append(F)
except Warning:
print("Warning encountered")
return
self.testset_dict[ref].append(
(selected_label, np.mean(F_avg, axis=0)))
def generate_features_with_ref_segments(self, selected_tuple, mean, std):
selected_label = selected_tuple[0]
selected_segment = selected_tuple[1]
s1 = np.array(selected_segment)
F_avg = []
for ref_segment in self.ref_segments[self.ref_label]:
s2 = np.array(ref_segment)
try:
F = feature_gen(s1, s2, mean, std)
F_avg.append(F)
except Warning:
print("Warning encountered..")
print("*************************")
return
#print(np.mean(F_avg, axis=0).shape)
self.trainset_list.append((selected_label, np.mean(F_avg, axis=0)))
np.save(
os.path.join(save_path, f"ref{len(self.ref_segments[0])}_dj{DJ}",
f"clf{self.ref_label}_2.npy"), self.trainset_list)