In short, BrainNetCNNs are convolutional neural networks for connectomes. CNNs are an example of neural networks for deep learning; a special kind of machine learning that attracted a lot of attention recently given its success in producing excellent results for a wide variety of prediction problems. Connectomes are brain networks that encode brain connectivity (structural or functional) as graphs with edges (connections) between nodes (brain regions). The strength of edges in the network may reflect the strength or confidence in the corresponding brain connection. BrainNetCNN is a framework for training a deep network to make predictions from connectomes. BrainNetCNN is optimized or trained on a set of connectomes with known output variables (e.g. clinical variable, e.g. disease state or performance score). The trained BrainNetCNN is then used to predict an output given a new connectome. BrainNetCNN introduced special connectome-driven layers to enable deep learning from connectomes. To delve into the details, take a look at our publications.