results = pd.DataFrame()
# sweep over the parameter lambda
for sigma in sigmas:
# reset total error counts for mean error metric
SFCLS_Error_Total = 0
FCLS_Error_Total = 0
LASSO_Error_Total = 0
p_norm_Error_Total = 0
delta_Error_Total = 0
# initialize experiment
np.random.seed(1)
random.seed(1)
endmembs = spec_lib("asu",
ascii_spectra="input/endmemb_libs/rogers_tes73.txt",
meta_csv="input/endmemb_libs/rogers_meta.csv")
synth = SynthMix(thresh=min_endmemb)
metrics = []
N = endmembs.spectra.shape[0] # number of spectral channels
M = endmembs.spectra.shape[1] # number of endmembers
# iterate over K mixtures
for i in range(K):
# create synthetic mixture
mixture = synth.mars_mix(endmembs)
mixture_noisy = mixture.perturb(method="gauss", deviation=sigma)
# SFCLS prediction
x, _ = SFCLS_unmix(endmembs.spectra,
mixture_noisy.spectra,
idx_73pnt = np.append(idx_73pnt, np.arange(64, 110))
channels = 73
# import dataframe of emissivities and metadata
print("loading data from:")
file_name = "TES_emissivity_0.csv"
df = pd.read_csv(file_path + file_name)
print(file_path + file_name)
for i in range(1, 7):
file_name = "TES_emissivity_{}.csv".format(i)
df = df.append(pd.read_csv(file_path + file_name))
print(file_path + file_name)
# import bedrock endmembers
endmembs = spec_lib(
"asu_hdf",
hdf_file="/home/john/datasets/asu/bedrock_lib_v5_73pnts_highSi.hdf")
bedrock_idx = endmembs.index
# append atmospheric endmembers
atmosphere = spec_lib("asu_hdf",
hdf_file="/home/john/datasets/asu/atmlib_73channels.hdf")
endmembs.append(atmosphere)
Ne = endmembs.spectra.shape[1]
# setup output
FCLS_abundance = np.zeros((int(Nx), int(Ny), Ne))
FCLS_RMS = np.zeros((int(Nx), int(Ny)))
FCLS_SD = np.zeros((int(Nx), int(Ny), Ne))
SFCLS_abundance = np.zeros((int(Nx), int(Ny), Ne))
SFCLS_RMS = np.zeros((int(Nx), int(Ny)))
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)