Monday, February 28, 2011

Are the Laws of Nature Markovian?

The present laws of physics are Markovian.  The state of the world at the current time step, t, depends only on the state of the world at the previous time step, t-1.  (If there happen to exist closed timelike curves in our world then general relativity suggests that this may not be true even in physics.)

But other sciences may propose laws which are nonMarkovian.  Biology (evolution) has the memories stored in DNA, cognitive science has long term and short term memories, economics may exhibit behavior (cooperation) that depends on a memory of past events, etc.

Saturday, February 19, 2011

Value pluralism again

It's reasonable to value patterns that are longer and which occur more frequently.  In Asa H lite Duration(A) is the length of pattern A and N(A) is the number of times pattern A has been seen.  But should these two values be given equal weight and combined as a product? i.e., should the utility of pattern A be U(A)=Duration(A)*N(A) as it is in Asa H lite?

An alternative is to keep Duration(A) and N(A) as components of a vector utility.  Instead of preferring pattern A to pattern B when U(A)>U(B) we might check to see if (Duration(A)>Duration(B) AND N(A)>N(B)) OR (Duration(A)>Duration(B) AND N(A)=N(B)) OR (Duration(A)=Duration(B) AND N(A)>N(B))

Thursday, February 10, 2011

An Asa H. 2.0 lite in BASIC

There are many versions (~100) of Asa H 2.0.  This is a lite version without many "bells and whistles" (time warping, image manipulation, etc.)
(Line 166 should read: Th3=.05)