|Jul 29, 2019|
|1 month 21 days ||Adding example program an data for the statistical filters.
|1 month 21 days ||Finalizing the documentation for KF & PF with two more examples.
KF: output error estimation as standard deviation, alias for
substitution '1st'. PF: memory bugfix for model fitting, less
particles as default.
|Jul 25, 2019|
|1 month 25 days ||Documentation for the new filters. It is reachable via 'Modules' but
not linked in the processing list yet. KF: filter only the param it
runs on by default, better nodata handling, system matrix can also
read a vector, cleanup. PF: bugfix when using…
|Jul 18, 2019|
|2 months 1 day ||PF: Reintroducing previously kicked idea to use data points as model
by means of fitting a curve. Alternative choice for most likely
path. System and observation model's substitutions variables don't
have to match anymore. KF: Output of evolution of…
|Jul 15, 2019|
|2 months 3 days ||The time Loop filter was using a different name than in its
documentation. It really is TimeLoop!
|2 months 3 days ||Slightly improved documentation and smarter warning when duplicate
timestamps are made of nodata values. And by the way, I did see the
previous commit :-)
|2 months 4 days ||Quick removal of duplicate code from the last commit before anyone
|2 months 4 days ||'Trim' overload to return strings. KF: Should now be well-defined,
first results have been achieved. Input matrices can hold brackets
and be built from diagonal vectors, initial state can be set
automatically from meteo data, better parse errors. PF…
|2 months 4 days ||Adding a missing documentation entry
|Jul 12, 2019|
|2 months 6 days ||Oups, my previous commit was a mess...
|2 months 6 days ||Added the possibility to add custom compiler flags (this should help
for issue 740) and to turn LTO off when compiling with optimizations
(for systems with little available RAM). Both are marked as
"advanced" and thus hidden from view for the normal…
|Jul 11, 2019|
|2 months 7 days ||PF & KF: nodata handling and adaptive timestepping; some
const-ness. KF: input of time-dependent system matrix, core fix (but
still not usable very well).
|Jul 10, 2019|
|2 months 9 days ||PF: Possibility to dump internal states to file system and read it
at next startup --> on successive calls the past does not have to
be recalculated and observations can be aggregated online. Smaller
Eigen include, parse_args fix. KF: Core algorithm…
|2 months 9 days ||The IDW_Slopes spatial interpolation needs to cache the data and was
failing to clear the data before the next run, leading to
accumulating more and more data (which is bad) and mixing data
between different calls (which is even worst). This has bee…
|2 months 9 days ||Safeguard in readLineToVec, Kalman filter: most input matrices can
be read, cleanup. Particle Filter: allowing to seed resampling RNG
for reproducability, allowing to set resampling threshold, cleanup.
No more GCC warnings.
|2 months 9 days ||In order to avoid Eigen's (tons of) warnings, it must be declared as
system library (as was already done) but also included without any
path (so it is handled as a system library)
|Jul 9, 2019|
|2 months 9 days ||Resampling fix and cosmetics in particle filter, started Kalman
filter, readLineToVec(delim) overload to return doubles instead of
|2 months 9 days ||The time loop filter had some limitations (regarding the definition
of the synchronization date) that have been removed. there was also
a small mistake that could lead to slightly sifted timestamps (ie
starting from the wrong one). This has also bee…
|Jul 8, 2019|
|2 months 11 days ||PF: Implemented paths resampling. With this, the filter should be
somewhat usable already.
|2 months 11 days ||Particle filter: Kicking idea to init via vector (if we allow that,
the user may aswell just write it themselves...), 3rd and last RNG,
2nd and last model function.