similar to previous gist, but draws 2D maps for each SVD components - good for analysis of 2D spectral mapping of samples

SVD2Dmap_for_plotcommander.py

#plotstyle = "basis"
plotstyle = "amplitude"
decimatex = 1
ncomponents = 6
a = ys[:, ::decimatex]
xs = xs[:, ::decimatex]
U, s, V = np.linalg.svd(a, full_matrices=False)

if plotstyle=="basis":
  for x, y, label in         zip(xs[:ncomponents], V[:ncomponents], range(ncomponents)):
    ax.plot(x, y, label="SVD component %d" % (label+1), lw=5*.6**label)
    ax.set_xlabel('wavelength (nm)')
    ax.set_ylabel('spectral intensity')
  ax.legend(loc='best', prop={'size':10})
  np.savetxt('svd-basis.txt', np.vstack([xs[0], V[:ncomponents]]).T, fmt="%.6f")

if plotstyle=="amplitude":
 print(U.shape, s.shape)
 res = (np.dot(U,np.diag(s)).T)
 ## assume the spectra were measured in a square NxN pattern:
 dim = int(len(res)**.5)
 subplotx,subploty=3,2
 fig.delaxes(ax)
 axs = fig.subplots(subploty,subplotx)
 axs = [ax for axx in axs for ax in axx] ## flatten
 print(enumerate(axs), axs)
 for compn,ax in enumerate(axs):
  print(compn)
  print(res[compn])
  component = res[compn].reshape(dim,dim)
  x,y = np.arange(dim), np.arange(dim)[::-1]
  levels = np.linspace(np.min(component), np.max(component), 50) 
  ax.contourf(x,y,component[::1][::-1], levels=levels, extend='both', cmap=cm.jet)
  ax.contour(x,y,component[::1][::-1], levels=levels, cmap=cm.jet) # avoid white stripes
  ax.set_xlim(x[0], x[-1])
  ax.set_ylim(y[0], y[-1])
  ax.set_title('SVD coefficient {}'.format(compn+1))
  np.savetxt('svd-map-of-coefficient{}.txt'.format(compn), (component).T, fmt="%.6f")
  np.savetxt('svd-map-of-coefficient{}_yflip.txt'.format(compn), (component[:][::-1]).T, fmt="%.6f")
  np.savetxt('svd-map-of-coefficient{}_xyflip.txt'.format(compn), component, fmt="%.6f")