The serotonergic system modulates brain processes via functionally distinctsubpopulations of neurons with heterogeneous properties, including theirelectrophysiological activity. In extracellular recordings, serotonergicneurons to be investigated for their functional properties are commonlyidentified on the basis of "typical" features of their activity, i.e. slowregular firing and relatively long duration of action potentials. Thus, due tothe lack of equally robust criteria for discriminating serotonergic neuronswith "atypical" features from non-serotonergic cells, the physiologicalrelevance of the diversity of serotonergic neuron activities results largelyunderstudied. We propose deep learning models capable of discriminating typicaland atypical serotonergic neurons from non-serotonergic cells with highaccuracy. The research utilized electrophysiological in vitro recordings fromserotonergic neurons identified by the expression of fluorescent proteinsspecific to the serotonergic system and non-serotonergic cells. Theserecordings formed the basis of the training, validation, and testing data forthe deep learning models. The study employed convolutional neural networks(CNNs), known for their efficiency in pattern recognition, to classify neuronsbased on the specific characteristics of their action potentials.