Permalink
Switch branches/tags
Nothing to show
Commits on Feb 15, 2017
  1. Adding the DOI badge.

    Guilherme Castelao
    Guilherme Castelao committed Feb 15, 2017
Commits on Apr 11, 2014
  1. Adding link to our paper.

    castelao committed Apr 11, 2014
Commits on Mar 8, 2014
Commits on Jun 17, 2013
  1. Including Acknowledgmets

    castelao committed Jun 17, 2013
  2. Ignoring *.h5

    castelao committed Jun 17, 2013
Commits on May 19, 2013
  1. Minor version upgrade.

    castelao committed May 19, 2013
  2. Fixing the drunken_drive()

    castelao committed May 19, 2013
Commits on May 16, 2013
  1. Including the function sampler()

    castelao committed May 16, 2013
    Part of the estimate_error, related to the actual sampling, was moved
      to a new function sampler().
Commits on May 15, 2013
  1. RingCenterFlex now works with (Lat,Lon) or (x,y).

    castelao committed May 15, 2013
    Once I migrated the basic system to the RingCenter class, I used the
      fundamental variables (t, x, y, u, v). Now the RingCenterFlex works
      with ( (t|datetime), ((Lat, Lon) | (x, y)), u, v).
Commits on May 13, 2013
  1. Organizing and bugfix to set ring's velocity structure.

    castelao committed May 13, 2013
    To estimate the distortion of the data xr, yr, I must use
    self.input['t'], not self.data['t'].
Commits on May 12, 2013
Commits on May 11, 2013
  1. Medium level upgrade. Now is RingCenter, instead of Ring.

    castelao committed May 11, 2013
    The center is now estimated by the class RingCenter, which expects the
    default mks input (t,x,y,u,v).
    
    The class RingCenterFlex will handle the LatxLon and datetime inputs.
Commits on May 9, 2013
Commits on May 7, 2013
  1. Heavy changes, migrating Ring to RingCenter.

    castelao committed May 7, 2013
    A new concept, RingCenter will be the base class, working on (t, x, y,
    u, v). And other classes will be build on the top, using that to handle
    other formats, like (datetime, Lon, Lat, u, v).
Commits on May 5, 2013
  1. I'm back to don't set initial p0 based on the dataset.

    castelao committed May 5, 2013
    I still think it has some future here, but need to work on it.
  2. Increasing the default lambda back to 1e-2.

    castelao committed May 5, 2013
    Looks like 1e-3 still allows some over-fitting.
Commits on Apr 25, 2013
  1. Don't like zero as initial guess.

    castelao committed Apr 25, 2013
    Temporary solution to don't let p0 starts with 0. The adjusting system
    don't like it.