Design of an optical content-addressable parallel processor for expert systems.

Area of Science:

• Computer Science
• Artificial Intelligence
• Optical Computing

Background:

• Rule-based systems (RBS) face performance limitations due to slow execution speeds.
• Existing electronic and optical systems for RBS acceleration have drawbacks, including time-consuming operations and underutilization of parallelism.

Purpose of the Study:

• To propose a novel optical content-addressable parallel processor for expert systems.
• To enhance the speed and efficiency of rule-based systems by addressing their core operational bottlenecks.

Main Methods:

• Developed a processor that executes the match, select, and act operations of RBS in a highly parallel manner.
• Utilized a two-dimensional data representation to leverage optical parallelism across RBS units.
• Incorporated a parallel optical match unit for 2D pattern matching and a novel single-step conflict-resolution algorithm in the select unit.

Main Results:

• The proposed processor extracts and exploits all possible parallelism within a RBS.
• The system efficiently processes general-domain knowledge, including variables, numbers, symbols, and comparison operators.
• Achieved high-speed execution of core RBS operations through optical parallel processing.

Conclusions:

• The optical content-addressable parallel processor offers a significant speed improvement for rule-based systems.
• The design is suitable for high-speed, general-purpose RBS applications.
• This approach overcomes the limitations of current electronic and optical systems for RBS acceleration.