文档介绍：Parallel Linear ic Programming
Carlton Downey and Mengjie Zhang
School of Engineering puter Science
Victoria University of Wellington, Wellington, New Zealand
Carlton.******@,
Mengjie.******@
Abstract. Motivated by biological inspiration and the issue of code dis-
ruption, we develop a new form of LGP called Parallel LGP (PLGP).
PLGP programs consists of n lists of instructions. These lists are exe-
cuted in parallel, after which the resulting vectors bined to pro-
duce program output. PGLP limits the disruptive effects of crossover and
mutation, which allows PLGP to significantly outperform regular LGP.
1 Introduction
Derived from ic algorithms [5], ic Programming (GP) [2, 6] is a promis-
ing and nature inspired approach to constructing reliable solutions to a range
of problems quickly and automatically, given only a set of human labeled in-
stances on which an evolved program can be evaluated. GP uses ideas analogous
to biological evolution to search the space of possible programs to evolve a good
program for a particular task. Since the 1990s, GP has been essful for solving
many machine learning problems [7, 9, 11–13].
In conventional GP, programs are trees of operators. This form of GP is
known as Tree based GP (TGP). Linear ic Programming (LGP) is an
alternative form of GP where individuals in the population are sequences of
instructions. Structuring programs as sequences of instructions has many ad-
vantages over structuring them as trees. LGP has been shown to significantly
outperform TGP on machine learning tasks such as multiclass classification [4,
8].
LGP performs well on multiclass classification problems because of the power
of its flexible program structure. LGP programs consist of a sequence of instruc-
tions which operate on a list of registers. This allows multiple outputs and per-
mits puted early in program execution to be reused later. These two
properties make LGP a powerful problem