Аннотация:The CHIME/FRB collaboration has recently published a catalog containing about half a thousand fast radio bursts (FRBs) including their spectra and several reconstructed properties, like signal widths, amplitudes, etc. We have developed a model-independent approach for a classification of these bursts using cross-correlation and clustering algorithms applied to one-dimensional intensity profiles of the bursts (i.e., to amplitudes as a function of time averaged over the frequency). Using this algorithm we identified two major classes of FRBs featuring different waveform morphology, and, simultaneously, different distribution of brightness temperature. Bursts from one of the identified cluster have lower brightness temperatures and larger widths than events from the other cluster. Both groups include bursts from the repeaters and one-off events.