Abstract: Parallel Split Learning (SL) allows resource-constrained devices that cannot participate in Federated Learning (FL) to train deep neural networks (NNs) by splitting the NN model into parts.