LINDA for case base retrieval: a case for extending the functionality of LINDA and its abstract machine

摘要

Case based reasoning (CBR) retrieval operations with partial ordering similarity measures have been implemented in LINDA to extract generic operations and identify potential parallelism in CBR. Mapping and reduction operations, and constraint matching versions of all retrieval operations were identified, then implemented in LINDA. Their operational semantics were expressed in terms of the CHAM for Liam (LINDA abstract machine) and refined to produce the basis for efficient, parallel implementations of the operations as additional LINDA instructions. The implementation of the constraint matching operations suggests the use of eval() to achieve process migration in distributed systems, and default local tuple space interaction. It as suggested that the constraint matching instructions replace the basic matching retrieval instructions, as they extend the functionality of the matching process by maintaining atomicity of matching and moving the constraint function from the program to the LINDA kernel.