Abstract
Identifying the computational roles of different neuron families is crucial for understanding neural networks. Most neural diversity is embodied in various types of γ-aminobutyric acid-mediated (GABAergic) interneurons, grouped into four major families. We collected datasets of opto-tagged neurons from all four families, along with excitatory neurons, from both the neocortex and hippocampus. The physiological features of these neurons were used to train a machine learning classifier, which subsequently inferred specific interneuron families in large-scale recordings. This combined approach enabled the reconstruction of synaptic connectivity motifs across interneuron family members. We further showed that these motifs differentially control the place field features of pyramidal neurons. Our findings attribute a prominent role to interneurons in the formation of a flexible cognitive map.