____                                       
/\  _`\                                     
\ \,\L\_\  __  __  __     __     __  __     
 \/_\__ \ /\ \/\ \/\ \  /'__`\  /\ \/\ \    
   /\ \L\ \ \ \_/ \_/ \/\ \L\.\_\ \ \_\ \   
   \ `\____\ \___x___/'\ \__/.\_\\/`____ \  
    \/_____/\/__//__/   \/__/\/_/ `/___/> \ 
                                     /\___/ 
                                     \/__/  

sway:

• verb
  • move or cause to move slowly or rhythmically backward and forward or from side to side. "he swayed slightly on his feet". synonyms: swing, shake, oscillate, undulate, move to and fro, move back and forth More
• noun
  • rule; control. "the part of the continent under Russia's sway". synonyms: clout, influence, power, weight, authority, control "his opinions have a lot of sway"

Magic parameters are stored in functions that start with an UPPER CASE name and are decorated by @setting. e,g

@setting
def CUT(): return o( 
  crowded=4, 
  cohen=0.1,
  fayyad=False
)

Once written that way, we say that the above numbers are the defaults for the global dictionary the.CUT.

These parameters can be accessed in the natural way:

the.CUT.cohen    # returns 0.1

The defaults can be altered as follows:

CUT(cohen=0.2)

These values can be reset to their initial defaults using:

 reset(seed=1)

This call finds all sets and sets them back to their defaults (as well as resetting the random number seed to "1").

Every file tests itself; e.g. to run scripts in boot.py:

python boot.py

The following files (that have check marks) currently are passing their tests:

• boot.py
• lib.py
• counts.py
• space.py
• grid.py
• mutate.py
• model.py
• models.py # tournament
• control.py

99% of all items in one cell.

Cells where objectives X and Y both get better may be in different locations.

"Good" cells (with more than N items) may be islands (not form 1 clump).

The IQR was not much reduced in bins than overall.

• Evaluate N items in a bin. Compute the mean values of the N objective in that bin.
• Select bins where the mean score of any objective is less the mean objective of all bins.

Two treatments:

1. Evaluate all items in all bins
2. Evaluate just the item near the middle of each bin.

Rx2 selects many of the same bins as selected by Rx1.

Given B*B bins, less than a dozen have more than 10 items and selected by the centrality works, ish method.