Antti's suggestion and a possible solution - online education

[Mark Tarver]

Antti Ylikoski (https://groups.google.com/forum/?hl=en#!topic/qilang/p-MrQ9hCiwU) suggested a whole bunch of interesting implementations.

Back in the day I wanted to implement a logic lab; a software playground for people to experiment with AI reasoning approaches.   The trouble as ever is funding.  

Now I'm going to finish an online course very soon and look to prepare the next one.  I found I could not do LPC as a course - the book made that rather redundant.   However what would people feel about a course on mechanised reasoning starting with databases, to semantic nets, Horn clause logic, FOL and high performance inferencing? 

Mark 

[Mark Tarver]

FYI - here is some output from a high-performance first-order logic engine I've just built in the last few days called HYPERPLAN.   HYPERPLAN has compiled the axioms of naive set theory (st) - (prod x y z) = x is the product of y and z, (eq x y) = x and y are the same sets; un = union.  It shows the kind of thing one can do in Shen.  As you can see even on a feeble monocore 1.3GHz laptop - you can get 1 million inferences a second.  This is the kind of stuff one might study.

Mark

 

(1-) (hyperplan (st))

compiling background theory

run time: 16.441999912261963 secs

----------------------------------------

enter assumptions> []

enter conclusion> [all x [all y [[prod x y e] => [eq x e]]]]

run time: 0.01599884033203125 secs

3133 inferences

true

----------------------------------------

enter assumptions> []

enter conclusion> [all x [all y [all z [[sub z y] => [[un x y z] => [un x y y]]]]]]

run time: 0.015001296997070313 secs

14161 inferences

true

----------------------------------------

enter assumptions> []

enter conclusion> [all x [all y [[pow x y] => [m y x]]]]

run time: 1.4039993286132813 secs

1469383 inferences

true

[Mark Tarver]

oh e = the empty set and pow x y = x is the powerset of y

[Antti Ylikoski]

I think that it would be a good idea.  However, I will have to delve into the LPC thoroughly before I can say anything very exact and very detailed about the activity.

A. J. Y.

[Antti Ylikoski]

https://en.wikipedia.org/wiki/List_of_artificial_intelligence_projects

perjantai 12. kesäkuuta 2015 23.25.04 UTC+3 Mark Tarver kirjoitti:

[Mark Tarver]

Hyperplan 1.02 - better performance.

(2-) (hyperplan (st))

compiling background theory

run time: 16.754000186920166 secs

----------------------------------------

enter assumptions> []

enter conclusion> [all x [all y [[prod x y e] => [eq x e]]]]

run time: 0.0 secs

2302 inferences

true

----------------------------------------

enter assumptions> []

enter conclusion> [all x [all y [all z [[sub z y] => [[un x y z] => [un x y y]]]]]]

run time: 0.01599884033203125 secs

18995 inferences

true

----------------------------------------

enter assumptions> []

enter conclusion> [all x [all y [[pow x y] => [m y x]]]]

run time: 0.0149993896484375 secs

18679 inferences

true

Shen Prolog is actually faster than I thought; for an occurs-check Prolog on a modest laptop, these are good figures.

Mark