System and method for autonomic optimization by computer programs

A method of optimizing a computer program, said method comprising:

• specifying at least one point of choice, determined by a programmer, in said computer program;
• defining a set of alternate choices at each point of choice, wherein said set of alternate choices include operational choices comprising:
• inputting a selection command that selects one function from a list of pre-selected functions for input into said computer program, wherein each function from said list of preselected functions is associated with a reward;
• allowing a learning protocol comprised of learning instructions in said computer program to track and reward said one function that is selected to determine an approximate optimal operation of said computer program based on said selection command;
• and setting at least one feedback point for said each point of choice;
• and wherein said set of alternate choices include operational choices, and wherein said method further comprises:
• inputting a reward command into said computer program at a point of choice, determined by a programmer, wherein said reward command provides reward in said computer program for said one function selected by said selection command, which results in said approximate optimal choice for optimizing said computer program.

A computer system comprising:

• a memory encoding a computer program;
• a pre-compiler that inputs a selection command at a point of choice, determined by a programmer, into a said computer program that runs on a computer, said selection command selecting one function from a list of pre-selected functions for input into said computer program, wherein each function from said list of pre-selected functions is associated with a reward;
• and a processor adapted to execute a learning protocol comprised of learning instructions in said computer program to track and reward said one function that is selected and determine an approximate optimal choice of operation of said computer program based on at least said selection command;
• and wherein said pre-compiler further inputs a reward command that provides a reward, at a point of choice, determined by a programmer, in said computer program, for said one function selected by said selection command, which results in said approximate optimal choice for self-optimizing said computer program.