DETECTION OF WEAK SIGNALS AGAINST BACKGROUND (NOISE) USING NEURAL-NETWORK CLASSIFIERS

We introduce a new neural classification technique for background reje ction in high energy physics and astrophysics experiments, which permi ts us to (i) directly optimize the desired quantity, i.e. the signific ance of signal detection (signal to noise ratio); (ii) obtain the comp licated nonlinear boundaries for signal event acceptance. Examples of implementing the proposed technique for background rejection in high e nergy astrophysics experiments are presented.