Figure 2.

1-d CNN for genomic prediction of a single trait with M SNP markers. The network has an input layer, two convolutional layers with their corresponding pooling layers, a fully connected hidden layer, and an output layer. Each convolutional layer applies a number of filters to the output of the previous layer, and its output is subsequently summarized by a pooling layer, where filters are arrays for convolving input. The number of filters generally increases as the network becomes deeper, and each filter learns a different abstract representation of the input data from a previous layer.