spacecruft
/
karoo_gp
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

2018 04/22

pull/13/head
kstaats 2018-04-22 13:58:39 -07:00
parent c6cee4928f
commit fc78518e9c
1 changed files with 56 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

57
RELEASE_NOTES.txt
View File

@ -1,6 +1,61 @@
2018 04/22
1) Completed the isolation of the user interface from the base_class such that all mid-run parameter configurations and
population summaries are now conducted in menu.pause(), and any associatede executions (Tree validation) are conducted
fully in the base_class.
karoo_gp.py imports karoo_gp_base_class.py
karoo_gp_base_class.py imports karoo_gp_pause.py
The user's initial run configuration is conducted in karoo_gp.py. Those settings are passed into karoo_gp_base_class.py
where they are used to guide the run. At each 'pause' (generation, interactive, or debug display modes; or end of run)
the user is taken into menu.pause where additional queries and configurations are conducted. Those value changes are
returned to the base_class where they are executed. When a run is completed (Server or Desktop), the user is returned
to the original karoo_gp.py for salutations and run termination.
Finally, the Karoo GP Python pasta is no longer spaghetti!
2) Modified the function of 'cont' such that it either continues if you yet have additional generations to go, or
querie the user if gen_id = gen_max and the prior run is done.
3) Replaced 'generation_id' with 'gen_id' and 'generation_max' with 'gen_max'.
4) Conducted a full set of unit tests on every user interface query in menu.pause().
5) Cleaned up the code beneath the user interface (menu.pause), making the while-true loops consistent with those in
karoo_gp_pause.py.
6) Merged 'fx_karoo_continue' into 'fx_karoo_gp' such that a simple while loop maintains an active evolutionary process
until the either the Server script is complete, or the user has terminated the run. This allows for the simple addition
of generations to an existing run. A far better solution than what I had previously coded.
7) Discovered that the loading of seed population_s is far too delicate. Even the slightest change and it fails. I have
disabled the load function for now, until I have time to rebuild it.
8) Please note that all modules are now stored in modules/
2018 04/19
Added a new kind of gene_pool filter (environmental pressure) such that you can select partial or full inclusion of
the operands (features). While not yet incorporated into the user interface (desktop or server), the new 'swim' setting
enables you to build consecutive gene pools that have *only* Trees that include at least one instance of each feature.
This may be helpful for regression problems moreso than classification, as regression tends to fall to a minimum
feature set, depending upon the nature of the solution (fixed or rolling value). I desire to merge tree_depth_min
setting and this new 'swim' into one user defined system that controls the gene_pool.
Discovered 3 more functions that did not include a "Called by: ___" statement. Updated.
Fixed a long-standing issue of total failure if you enter 1 for the total number of generations. This setting, while
it may seem odd, allows Karoo GP to be used to create populations of stochastic multivariate expressions based upon the
given dataset for use in other applications. Simple fix. The population is written to the runs/ directory as with all
other modes of operation, including Play.
2018 04/12c
Updated the User Guide and modified a help print statement in the sys args of the Server function.
Updated the User Guide to match the Desktop and Serve merger, and modified a help print statement in the sys args of
the Server function.
2018 04/12b