The problem of tuning weights/classifying feature vectors is well known problem in CBR. Previously neural nets have been successfully incorporated in many applications associated with knowledge-based systems. Knowledge acquisition, feature selection, classifying feature vectors, tuning the weights is some of those many avenues in CBR where neural nets can be applied. This paper first describes some related previous work to learn feature weights. After that, this tries to see the applicability of neural nets to the current problem, formulates the problem's solution using a neural net topology. Next, this articulates an algorithm to be used in conjunction with Back propagation and tries to give theoretical foundation to the proposed algorithm. Then, this tries to establish the effectiveness of the proposed solution by running experiments on some critical sample domains, and results are published and studied. Finally, this accepts the challenges in the proposed integration work along with other bottlenecks and the future work expected in this direction.
[1]
Janet L. Kolodner,et al.
Case-Based Reasoning
,
1989,
IJCAI 1989.
[2]
Michel Manago,et al.
INRECA: A Seamlessly Integrated System Based on Inductive Inference and Case-Based Reasoning
,
1995,
ICCBR.
[3]
Deepak Khemani,et al.
InfoFrax : CBR in Fused Cast Refractory Manufacture
,
2002,
ECCBR.
[4]
I C G Campbell,et al.
Constructive learning techniques for designing neural network systems
,
1998
.
[5]
Kalyan Moy Gupta,et al.
A connectionist approach for similarity assessment in case-based reasoning systems
,
1997,
Decis. Support Syst..
[6]
Jacek M. Zurada,et al.
Introduction to artificial neural systems
,
1992
.
[7]
Padraig Cunningham,et al.
Using Introspective Learning to Improve Retrieval in CBR: A Case Study in Air Traffic Control
,
1997,
ICCBR.