To compile the C++ implementation the following dependencies have to be installed:
- GNU Scientific Library
- Atlas and Blas
An example for using the code with C++ can be found under examples.
If you also want to compile and use the Python interface you need the following extra packages
- Swig
- Numpy
To visualize the result you may also want to install matplotlib.
Load the required packages
import gba
import numpy as np
Set up some filterbank input data and the magnitude and distance samples for sampling the pdf:
m = np.linspace(2, 8, 200)
r = np.linspace(0,2.0,200)
fb_data = np.array([3.8221, 6.7219, 5.6877, 14.739,22.047, 31.288, 15.537,
3.8673,1.2311])*1e-4
pdf = np.zeros((m.size,r.size))
m_marginal = m.copy()
r_marginal = r.copy()
Estimate the pdf:
# Number of most similar traces in training data
nsim = 30
# Number of filter bands
nbands = 9
# Time of the filterbank observation since the P-wave detection
time_since_p = 3.0
g = gba.GbA(nbands,nsim)
# Load the training data
g.init()
# Set the magnitude and distance samples
g.set_samples(m,r)
# Compute mean and cross-covariance matrix
g.process(fb_data,time_since_p,gba.TData.vertical)
g.get_pdf(pdf,m_marginal,r_marginal)
Plot the result
import matplotlib.pyplot as plt
from matplotlib.ticker import NullFormatter
fig = plt.figure()
# definitions for the axes
left, width = 0.1, 0.65
bottom, height = 0.1, 0.65
bottom_h = left_h = left + width + 0.02
rectpdf = [left, bottom, width, height]
rectr = [left, bottom_h, width, 0.2]
rectm = [left_h, bottom, 0.2, height]
# start with a rectangular Figure
fig = plt.figure(1, figsize=(8, 8))
axpdf = fig.add_axes(rectpdf)
axm = fig.add_axes(rectm)
axr = fig.add_axes(rectr)
axpdf.contourf(r, m, pdf)
axpdf.set_xticks([np.log10(x) for x in [5,10,20,50]])
axpdf.set_xticklabels([5,10,20,50])
axpdf.grid()
# plot marginals
axm.plot(m_marginal,m)
axr.plot(r,r_marginal)
mhat = m[np.argmax(m_marginal)]
rhat = r[np.argmax(r_marginal)]
axpdf.vlines(rhat,m.min(),m.max(),color='k')
axpdf.hlines(mhat,r.min(),r.max(),color='k')
axpdf.set_xlabel('Distance [km]')
axpdf.set_ylabel('Magnitude')
# no labels
nullfmt = NullFormatter()
axm.yaxis.set_major_formatter(nullfmt)
axm.xaxis.set_major_formatter(nullfmt)
axr.yaxis.set_major_formatter(nullfmt)
axr.xaxis.set_major_formatter(nullfmt)
