def plotrefdepx(axfiles, lam):
bands = read_data.getband()
fontsize = 18
matplotlib.rcParams.update({'font.size': fontsize})
fig = plt.figure(figsize=(7, 3))
ax = fig.add_subplot(111)
lss = ["dashed", "dashed", "dashed", "solid", "solid", "dotted", "dotted"]
for i in [0, 4, 6]:
axfile = axfiles[i]
A, X, resall = read_data.readax(axfile)
print(lam[i])
plref_each(X, bands, lss[i], "$\lambda_X=10^{" + str((lam[i])) + "}$")
plt.ylabel("Unmixed Spectra", fontsize=16)
plt.xlabel("wavelength [micron]", fontsize=16)
plt.legend(fontsize=13)
plt.savefig("refdep.pdf", bbox_inches="tight", pad_inches=0.0)
plt.show()
def plot_rega(axfiles):
NN = 2435 * 7.
# NN=1
lcmean = 107.40646786191814
mrarr = []
mrsaarr = []
likarr = []
detxn = []
for i, axfile in enumerate(axfiles):
A, X, resall = read_data.readax(axfile)
Aclass = diffclass.mclassmap(A)
mr = diffclass.classdif(Aclass)
AFnorm = np.sqrt(np.sum(A * A))
mrarr.append(mr)
likarr.append(resall[-1][1])
Xn = np.copy(X)
for k in range(0, np.shape(X)[0]):
Xn[k, :] = X[k, :] / np.sum(X[k, :])
detxn.append(np.linalg.det(np.dot(Xn, Xn.T))) #/10**lam[i])
likarr = np.array(likarr)
detxn = np.array(detxn)
fontsize = 18
matplotlib.rcParams.update({'font.size': fontsize})
fig = plt.figure(figsize=(7, 5))
ax = fig.add_subplot(211)
ax.plot(10**lam, np.sqrt(likarr / NN) / lcmean, "o", color="C0")
ax.plot(10**lam, np.sqrt(likarr / NN) / lcmean, color="C0")
plt.xscale("log")
plt.ylabel("mean residual")
ax = fig.add_subplot(212)
plt.xscale("log")
# plt.yscale("log")
# ax.plot(10**lam,(absa),"o",)
ax.plot(10**lam, (detxn), "o", color="C0")
ax.plot(10**lam, (detxn), color="C0")
plt.ylabel("$\det{(\hat{X} \hat{X}^T)}$")
plt.xlabel("$\lambda_A$")
plt.savefig("regad.pdf", bbox_inches="tight", pad_inches=0.1)
plt.show()
def plotrefdirect(axfiles, lam):
bands = read_data.getband()
fontsize = 18
matplotlib.rcParams.update({'font.size': fontsize})
fig = plt.figure(figsize=(8, 5.5))
ax = fig.add_subplot(111)
cloud, cloud_ice, snow_fine, snow_granular, snow_med, soil, veg, ice, water, clear_sky = io_refdata.read_refdata(
"/home/kawahara/exomap/sot/data/refdata")
nnl = 1 #len(np.median(bands,axis=1))
u, val, normvveg = pm.norm(veg)
# ax.plot(u,val,c="gray",lw=2,label="vegitation (deciduous)")
ax.plot(u, val, c="gray", lw=2)
u, val, normvsoil = pm.norm(soil)
ax.plot(u, val, c="gray", lw=2, ls="dashed")
# ax.plot(u,val,c="gray",lw=2,ls="dashed",label="soil")
u, val, normvwater = pm.norm(water)
ax.plot(u, val, c="gray", lw=2, ls="-.")
# ax.plot(u,val,c="gray",lw=2,ls="-.",label="water")
lss = ["dashed", "solid", "dotted", "dashdot"]
sy = [False, True, False]
for i in range(0, 3):
axfile = axfiles[i]
A, X, resall = read_data.readax(axfile)
mmrsa = mrsa.mrsa_meanX(X)
print(mmrsa)
print(lam[i])
metric.plref_each(X,
bands,
lss[i],
"$\lambda_X=10^{" + str(int(lam[i])) + "}$",
symbol=sy[i])
plt.ylim(0, 0.6)
plt.ylabel("Unmixed Spectra", fontsize=16)
plt.xlabel("wavelength [micron]", fontsize=16)
plt.legend(fontsize=13)
plt.savefig("refdirect.pdf", bbox_inches="tight", pad_inches=0.0)
plt.show()
# ax.plot(pred[:,i])
plt.savefig("lc.pdf", bbox_inches="tight", pad_inches=0.0)
plt.show()
if __name__ == '__main__':
import sys
# axfile="npz/T116/T116_L2-VRLD_A-2.0X4.0j99000.npz"
# theme="dscovr"
theme = "3c"
axfile = sys.argv[1]
try:
title = sys.argv[2]
except:
title = ""
A, X, resall = read_data.readax(axfile)
lcmean = 105.45194310301967
lcsig = lcmean * 0.01
print("Ln L=", resall[-1, :] / (2 * lcsig * lcsig))
print("AIC=", 2 * resall[-1, :] / (2 * lcsig * lcsig) + 2 * (3072 * 3))
sys.exit()
bands = read_data.getband()
# bands=[[0.388,0.388],[0.443,0.443],[0.552,0.552],[0.680,0.680],[0.688,0.688],[0.764,0.764],[0.779,0.779]] #DSCOVR
# showpred(A,X)
# sys.exit()
fontsize = 18
matplotlib.rcParams.update({'font.size': fontsize})
title = ""
unwrap=np.array(unwrap).T
print(np.shape(unwrap))
return unwrap
def projpc(Xin,Xp):
pc1=Xp[0,:]
pc2=Xp[1,:]
xpc1=np.dot(Xin,pc1)
xpc2=np.dot(Xin,pc2)
return xpc1,xpc2
if __name__=='__main__':
import sys
axfile="npz/T215/T215_N3_L2-VRDet_A-1.0X2.0j100000.npz"
A,X,resall=read_data.readax(axfile)
W=np.load("w512.npz")["arr_0"]
WA=np.dot(W,A)
lcall=np.load("lcallN0.01.npz")["arr_0"]
for i in range(0,np.shape(lcall)[1]):
lcall[:,i]=lcall[:,i]/np.sum(WA,axis=1)
# lcall[:,i]=lcall[:,i]/np.sum(Ad,axis=1)
pca = PCA(n_components=2)
pca.fit(lcall) #x(k,l)
Xp=((pca.components_))
xpc1,xpc2=projpc(X,Xp)
lcpc1,lcpc2=projpc(lcall,Xp)
def plot_regx(axfiles):
NN = 512 * 10
lcmean = 105.70907195459881
mrarr = []
likarr = []
detx = []
detxn = []
absa = []
cpr = []
for i, axfile in enumerate(axfiles):
A, X, resall = read_data.readax(axfile)
Aclass = diffclass.mclassmap(A)
mr = diffclass.classdif(Aclass)
cpr.append(mr)
mmrsa = mrsa.mrsa_meanX(X)
print("mmrsa=", mmrsa)
AFnorm = np.sqrt(np.sum(A * A))
mrarr.append(mmrsa)
likarr.append(resall[-1][1])
absa.append(resall[-1][2])
Xn = np.copy(X)
for k in range(0, np.shape(X)[0]):
Xn[k, :] = X[k, :] / np.sum(X[k, :])
detxn.append(np.linalg.det(np.dot(Xn, Xn.T))) #/10**lam[i])
detx.append(np.linalg.det(np.dot(X, X.T))) #/10**lam[i])
print(lam[i], resall[-1][1], AFnorm, mmrsa)
likarr = np.array(likarr)
absa = np.array(absa)
detx = np.array(detx)
detxn = np.array(detxn)
cpr = np.array(cpr)
fontsize = 18
matplotlib.rcParams.update({'font.size': fontsize})
fig = plt.figure(figsize=(7, 9))
ax = fig.add_subplot(411)
ax.plot(10**lam, np.sqrt(likarr / NN) / lcmean, "o", color="C0")
ax.plot(10**lam, np.sqrt(likarr / NN) / lcmean, color="C0")
plt.ylabel("mean residual")
plt.xscale("log")
ax = fig.add_subplot(412)
plt.xscale("log")
# plt.yscale("log")
# ax.plot(10**lam,(absa),"o",)
ax.plot(10**lam, (detxn), "o", color="C0")
ax.plot(10**lam, (detxn), color="C0")
# ax.plot(10**lam,(detxn)*1000,color="C0")
plt.ylabel("$\det{(\hat{X} \hat{X}^T)}$")
# plt.ylabel("$\det{({X} {X}^T)}$")
# plt.ylabel("$||D - W A X||_F^2$")
ax = fig.add_subplot(413)
ax.plot(10**lam, mrarr, "o", color="C1")
ax.plot(10**lam, mrarr, color="C1")
plt.xscale("log")
plt.ylabel("$\overline{MRSA}$")
plt.ylim(0.03, 0.1)
ax = fig.add_subplot(414)
ax.plot(10**lam, cpr, "o", color="C1")
ax.plot(10**lam, cpr, color="C1")
plt.xscale("log")
plt.ylabel("CPR")
plt.xlabel("$\lambda_X$")
plt.savefig("regx.pdf", bbox_inches="tight", pad_inches=0.1)
plt.show()