search

CSTR-Edinburgh / merlin

This is now the official location of the Merlin project.
http://www.cstr.ed.ac.uk/projects/merlin/
Apache License 2.0
1.31k stars 440 forks source link

coarse coding features #488

Open fxwderrick opened 5 years ago

fxwderrick commented 5 years ago

Three coarse coding features vectors are exactly the same. `def compute_coarse_coding_features(self, num_states): assert num_states == 3

    npoints = 600
    cc_features = numpy.zeros((num_states, npoints))

    x1 = numpy.linspace(-1.5, 1.5, npoints)
    x2 = numpy.linspace(-1.0, 2.0, npoints)
    x3 = numpy.linspace(-0.5, 2.5, npoints)

    mu1 = 0.0
    mu2 = 0.5
    mu3 = 1.0

    sigma = 0.4

    cc_features[0, :] = mlab.normpdf(x1, mu1, sigma)
    cc_features[1, :] = mlab.normpdf(x2, mu2, sigma)
    cc_features[2, :] = mlab.normpdf(x3, mu3, sigma)

` Samples and mean are shifted in same manner. Variance remains the same thus three probability vectors are exactly the same.