-
Notifications
You must be signed in to change notification settings - Fork 1
/
run_labmix.py
227 lines (186 loc) · 7.48 KB
/
run_labmix.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
"""
Experiment to predict abundances from laboratory mixtures
from the Feely dataset
inputs: input
"""
from mixclass import MixClass
from endmember_class import spec_lib
from unmix import FCLS_unmix, inftyNorm_unmix, LASSO_unmix, pNorm_unmix, deltaNorm_unmix
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import os
import datetime
# experiment parameters
seed = 1
K = "N/A"
sigma = "N/A"
thresh = 0.01
run_FCLS = True
run_inftyNorm = True
run_LASSO = True
run_pnorm = True
run_delta = False
# algorithm hyperparameters
lam_LASSO = 1e-4
lam_infty = 1e-6
lam_delta = 1e-6
lam_pnorm = 1e-3
delta = 1e-2
p = 0.99
def main():
# start from a random seed
np.random.seed(seed)
# initialize experiment
endmembs = spec_lib("asu",
ascii_spectra="input/kim/kim_library_tab.txt",
meta_csv="input/kim/kim_library_meta.csv")
mixtures = MixClass(ascii_spectra="input/kim/mtes_kimmurray_rocks_full_tab.txt",
meta_csv="input/kim/mtes_kimmurray_rocks_full_meta.csv")
# crop spectra at 400 wavenumbers
endmembs.spectra = endmembs.spectra[104:, :]
mixtures.spectra = mixtures.spectra[104:, :]
endmembs.bands = endmembs.bands[104:]
metrics_FCLS = []
metrics_inftyNorm = []
metrics_LASSO = []
metrics_pnorm = []
metrics_delta = []
# iterate over mixtures
for i in range(len(mixtures.names)):
# extract next mixture
mixture = mixtures.single(i)
idx_pos_truth = np.nonzero(mixture.presence > 0)[0]
# check if mixture is labelled as valid
if mixture.category == "valid_mixture":
# spectral unmixing
if run_inftyNorm:
# SFCLS prediction
x = inftyNorm_unmix(endmembs.spectra, mixture.spectra, lam=lam_infty)
metrics_inftyNorm.append(Metrics(x, mixture, mixture, endmembs.spectra, idx_pos_truth))
if run_FCLS:
# FCLS prediction
x = FCLS_unmix(endmembs.spectra, mixture.spectra)
metrics_FCLS.append(Metrics(x, mixture, mixture, endmembs.spectra, idx_pos_truth))
if run_LASSO:
# LASSO prediction
x = LASSO_unmix(endmembs.spectra, mixture.spectra, lam=lam_LASSO)
x = x / sum(x)
metrics_LASSO.append(Metrics(x, mixture, mixture, endmembs.spectra, idx_pos_truth))
if run_pnorm:
# p-norm prediction
x = pNorm_unmix(endmembs.spectra, mixture.spectra, lam=lam_pnorm, p=p)
metrics_pnorm.append(Metrics(x, mixture, mixture, endmembs.spectra, idx_pos_truth))
if run_delta:
# delta norm prediction
x = deltaNorm_unmix(endmembs.spectra, mixture.spectra, lam=lam_delta, delta=delta)
metrics_delta.append(Metrics(x, mixture, mixture, endmembs.spectra, idx_pos_truth))
# save metrics
result_path, today = create_directory()
# create experiment index
i = 0
while os.path.exists(result_path + today + "_experiment_params%d.csv" % i):
i += 1
# store experiment parameters to export with results
experiment_params = [{"seed": seed,
"K": K,
"sigma": sigma,
"thresh": thresh,
"run_FCLS": run_inftyNorm,
"run_LASSO": run_LASSO,
"run_pnorm": run_pnorm,
"run_delta": run_delta,
"lam_LASSO": lam_LASSO,
"lam_delta": lam_delta,
"lam_pnorm": lam_pnorm,
"delta": delta,
"p": p}]
saveListOfDicts(experiment_params, "labmix_params", result_path, today, i)
# save results of experiments
if run_inftyNorm:
saveListOfDicts(metrics_inftyNorm, "labmix_infty", result_path, today, i)
if run_delta:
saveListOfDicts(metrics_delta, "labmix_delta", result_path, today, i)
if run_FCLS:
saveListOfDicts(metrics_FCLS, "labmix_FCLS", result_path, today, i)
if run_LASSO:
saveListOfDicts(metrics_LASSO, "labmix_LASSO", result_path, today, i)
if run_pnorm:
saveListOfDicts(metrics_pnorm, "labmix_pnorm", result_path, today, i)
def consolidate(abundances):
"""
consolidate endmember abundances into mineral abundances
to compare ground truth (mineral category) with prediction (endmembers)
"""
proportions = []
# amphiboles
proportions.append(abundances[0:3].sum())
# consolidate biotite-chlorite
idx = [3, 5, 23]
proportions.append(abundances[idx].sum())
# consolidate calcite-dolomite
idx = [4, 6]
proportions.append(abundances[idx].sum())
# epidote
proportions.append(abundances[9])
# feldspar
proportions.append(abundances[8:16].sum())
# garnet, obsidian, glaucophane, kyanite, muscovite
for j in [16, 17, 18, 19, 20]:
proportions.append(abundances[j])
# olivine
proportions.append(abundances[21:23].sum())
# pyroxene
proportions.append(abundances[24:30].sum())
# quartz
proportions.append(abundances[30:32].sum())
# serpentine
proportions.append(abundances[[32, 37]].sum())
# talc, vesuvianite, zoisite
for j in [33, 34, 35]:
proportions.append(abundances[j])
# normalize proportions by blackbody abundance
proportions = [p / (1 - abundances[36]) for p in proportions]
proportions = np.asarray(proportions)
return proportions
# Define metrics, computed for all algorithms using this function
def Metrics(x, mixture, mixture_noisy, em_spec, idx_pos_truth):
N = len(x)
recon = np.matmul(em_spec, x)
x = consolidate(x)
x_presence = np.zeros(x.shape)
x_presence[np.nonzero(x >= thresh)] = 1
precision = sum(mixture.presence[idx_pos_truth] == x_presence[idx_pos_truth]) / sum(x_presence)
recall = sum(mixture.presence[idx_pos_truth] == x_presence[idx_pos_truth]) / sum(mixture.presence)
metrics = {"RMS_true": ((1.0 / N) * sum((mixture.spectra - recon) ** 2)).item(),
"Error_L1": (sum(abs((x - mixture.proportions.transpose())))).item(),
"Error_L2": np.sqrt(sum((x - mixture.proportions.transpose()) ** 2)).item(),
"accuracy": sum(mixture.presence == x_presence) / len(mixture.presence),
"precision": precision,
"recall": recall}
return metrics
def create_directory():
now = datetime.datetime.now()
# create directory for today's date
today = now.strftime("%Y-%m-%d")
todays_results = "results/" + today
try:
os.mkdir(todays_results)
except OSError:
print("Creation of the directory %s failed" % todays_results)
else:
print("Successfully created the directory %s " % todays_results)
result_path = todays_results + "/"
return result_path, today
def saveListOfDicts(metrics, name, result_path, today, i):
"""
:param metrics: list[dict1, dict2, ...]
:param method: string, name of method
:return:
"""
metrics = pd.DataFrame(metrics)
print(name)
print(metrics.mean())
metrics.to_csv(result_path + today + "_%s%s.csv" % (name, i))
if __name__ == '__main__':
main()