CLASSIFICATION ALGORITHMS APPLIED TO 10588 LOW-RESOLUTION SLITLESS SPECTRA - A SEARCH FOR UNUSUAL AND INTERESTING OBJECTS

We have performed a search for unusual objects among nearly 11 000 low-resolution spectra taken from 23 grens plates of the CFHT-Laval survey (14.5 deg2; 14 < V < 18). The classification method used is based on a hierarchical clustering analysis technique. In a first step, we exploit the highly effective and objective power of the method at isolating uncommon energy distributions. The clustering analysis also provides us with a tool to determine synoptic groups for common stars in the sample. This permits us to identify peculiar objects using a template matching technique. We furnish positions, magnitudes, grens energy distributions and finding charts for 4 objects having energy distributions significantly different from the rest of the population. We identify 32 other objects of interest, among them: 14 strong emission line objects, 8 suspected degenerate stars and 3 very blue spectra without a Balmer discontinuity.