โ Scribed by B. Dubuisson
No coin nor oath required. For personal study only.
Patternrecognition is a welladaptedset of methods in orderto solve diagnostic problems. One of the main problemof diagnosis is to have a set of data corresponding to thedifferentpossible system faults in order to learn the decision rule. This set being not generally complete, we propose a decision rule basedon the conceptof reject: this ruledoes notneeda complete learning setat the beginning because the learning is done iteratively when observing the physical system. Thereject conceptis divided into two possibilities: ambiguity reject and distance reject. Each one is adapted to a different situation. Wedevelop main differences between these two rejects and exhibit theirinterests. At last, some methods are given aboutremoving a rejectdecision.
๐ SIMILAR VOLUMES
We have developed a high-speed pattern recognition system using evolvable hardware (EHW). EHW is hardware that can change its own structure by genetic learning for maximum adaptation to the environment. The purpose of the system is to show that recognition devices based on EHW are possible and that
Practical models used in identification of process control processes must be too simplistic to give precise control information. However, these models can be used for adaptation if they are continuously readapted. But the identification then lacks the precision which might justify the analytic elabo