To better classify human and mouse reads, we tuned several hyperparameters including kernel size, filters numbers, pooling methods and optimization algorithm to produce the highest accuracy (Supplemental Fig. 2). First, we noticed that the number of epochs with the highest accuracy turned out to be five, which is smaller than the usual number required in image processing or text classification. The accuracy on training sets increased dramatically after five epochs and continued increase to near 100% while the accuracy on testing sets dropped after five iterations (Supplemental Fig. 2A). This indicated a quick overfitting of the model63, 64 and signaled that the number of epochs should be set strictly within six.65–67 However, variations in the kernel size and number of filters, which are two critical parameters that alter network efficiency,68, 69 did not change the classification accuracy dramatically (Supplemental Fig. 2B–C). Kernel size only affected the overall performance of the model by around 1%. With a kernel size larger than four, the model stably produces accuracies over 87% (Supplemental Fig. 2B–C). Similarly, filter number only slightly changed the accuracy (Supplemental
