v2.64.0 beam algorithm

v2.64.0 - Add Test Algorithm 'beam':

This algorithm optimizes the deck based on a the normal (hill) climb algorithm, but uses multiple decks. Therefore this algorithm is less likely to produce local optima.

example usage:

./tuo gauntlet1 gauntlet2 beam 100 surge

Algorithm:

A simplified description of the algorithm:

Say we have a beam-size of 5. Then the algorithm always saves the 5 known best decks. Instead of trying to further optimize only the best deck, like hill climb does, beam-climb will test each of the 5 decks for new best performing decks. This procedure is repeated until no better decks are found.

Read Further: https://en.wikipedia.org/wiki/Beam_search

Comparison to climb:

• The algorithm is generally slower than normal climb, but yields better results.
• Best way to use this algorithm probably is to check an already simmed deck for final improvements.

Parameters:

• 'beam-size' sets the size of the beam/decks (default=5)

Implementation notes:

• You can pass gauntlets as your deck, this will then sim each of them and keep the best X of them (X='beam-size')
• If 'beam-size' excedes the number of passed decks, it will be filled with mutated decks (see Genetic Algorithm)

Caution:

• Default parameters are not perfect yet, playing around with the parameters might increase/decrease performance drastically (any feedback or ideas on changes are as always much appreciated).
• The algorithm should work with most tuo flags/modes, but not all are tested.
• Better check if the results seem reasonable

Other Notes:

• 'climb_forts' and 'genetic' bug fixes

Special Thanks to Ender Cloud for mentioning and explaining the algorithm.