Fig. 1

Topology of the deep convolutional visual encoding model. The visual stimulus input matrix passes through a 1D convolution process, followed by a maximum pooling stage then a flattening stage to produce a 1D visual feature vector. The firing history input matrix passes through similar stages to produce a 1D firing feature vector. The two feature vectors are concatenated to produce a combined feature vector which is presented to a fully connected neural network to finally predict firing rate values for all neurons