1.2-7 (16 Sept 2013)
-----
added variance of mean and DoF outputs to predict_R for
regression models
correction of title and description in relevance doc file
implementing qEntropy and other active learning stats
for learning about contours
fixed bug in tree proposals when tree prior a=0
added coef.dynaTree function to extract coefficients from
linear leaf models
added average number of leaves to treestats
1.2-6 (21 April 2012)
-----
dealt with NULL issues and updated to mystdout and
mystderr
1.2-5 (06 Jan 2012)
-----
added NAMESPACE file
1.2-4 (18 Dec 2011)
-----
now depends on R 2.14.0
1.2-3 (16 Dec 2011)
-----
added NDEBUG to Makevars
Fixed bug when retiring with forgetting in the linear model
(lambda < 1). Needed to re-calc bb and other sufficient stats
that don't need to be done in the lambda=1 case.
added treestats.dynaTree function and changed varprop*
functions to be methods of type dynaTree
implemented sameleaf function to calculation average number
of other points in the same leaf as each row of X
extra check of bb <= 0 in calculate_linear defaulting to
constant model, but there are still cases where this doesn't
work quite right
added Tree::Collapse in as a crude initial attempt to fix
trees with leaf nodes that become improper after a retirement
with forgetting
1.2-2 (27 Sept 2011)
----
Fixed bug in retire.dynaTree that was causing y-labels to be
off by 1 for classification; this was leading to an assertion
failure in sens.dynaTree
fixed typos in documentation
1.2-1 (2 Sept 2011)
----
Fully documented sensitivity analysis code and relevance
calculation code, IECI and other online learning code not
yet documented
1.2 (25 June 2011)
---
Fully documented the online learning additions and
added demos and examples
1.1 (13 May 2011)
---
added copy.dynaTree function using copy_cloud_R and
rejuvinate.dynaTree using rejuvinate_R; these are
still undocumented
updated JASA reference and changed to Booth addresses
discovered/fixed bug in classification marginal
likelihood
initial implementation of sensitivity analysis code
initial implementation of online retiring, and added
alcX and entropyX calculations for choosing retires,
and drift for historical retirement
multiple growing rectangle proposing types
specify Xref locations in ALC calculation, and analytic
version by specifying a rectangle through Xref
propagate posterior probabilities now in log space
implemented residual resampling for resample step
streamlined prediction by accumulating (summing) each
particle's predictive rather than storing in big
matrices which are then averaged
implemented predictive probability calculations for
Expected Likelihood comparisons
1.0-2 (10 Aug 2010)
-----
simplified the prediction routines under the LM with
centering
also now allows icept = FALSE with LM
1.0-1 (16 July 2010)
-----
changed to LU version of the hyper-rectangle for proposing
grow tree operations
added splitmin and basemax and added proj argument to
plot.dynaTree for 1-d projections
1.0 (02 June 2010)
---
Initial version