Volume 14, Number 3
PATTERN RECOGNITION LETTERS
In the 1960's we heard such comments about statistical versus, structural-linguistic techniques. What is incumbent on us is to attempt to understand the capabilities and applicability of the various tools and exploit the complementary advantages of the different paradigms. Examples of such attempts on combining linguistic and statistical models, combining heuristic search with statistical pattern recognition, and heuristic search with structural pattern recognition appear in Kanal and Chandrasekaran (1972), Kulkarni and Kanal (1976), Kanal (1979), Stockman and Kanal (1983), Kanal and Dattatreya (1986). A variety of combined approaches to the design of decision trees is surveyed in Dattatreya and Kanal (1985). Combining structural and probabilistic knowledge with hypergraph search to develop alternate scenarios for decision making in uncertain domains is presented in Bhatnagar and Kanal (1993a,b).
Figure 1 displays some of the tools currently available in the pattern recognition tool kit, and some of the problems to which these tools have been applied. In "Hybrid Systems-A Key to Intelligent Pattern Recognition" (Kanal and Raghavan (1992)), we discuss some of the key issues which need to be addressed in integrating
Figure 1 Some tools for a pattern recognition tool kit.
Figure 2. A hybrid of neural networks for RCS-based target recognition