Identification of causal relationships of neural activity is one of the most important problems in neuroscience and neural engineering. We show that a novel deep learning approach using a convolutional neural network to model output neural spike activity from input neural spike activity is able to achieve high correlation between the predicted probability of spiking in the output neuron and the true probability of spiking in the output neuron for data generated with a generalized linear model. The convolutional neural network is also able to recover the true model variables (kernels) used to generate the probability of spiking in the output neuron. Based on the convolutional neural network model's validation via a generalized linear model, future work will include validation with non-linear models that use higher-order kernels.