> A temporally convolutional DNN with five to eight layers was required to capture the I/O mapping of a realistic model of a layer 5 cortical pyramidal cell (L5

> A temporally convolutional DNN with five to eight layers was required to capture the I/O mapping of a realistic model of a layer 5 cortical pyramidal cell (L5PC). > When NMDA receptors were removed, a much simpler network (fully connected neural network with one hidden layer) was sufficient to fit the model.