Fig. 2From: Class similarity network for coding and long non-coding RNA classificationThe comparison between Class Similarity Network and baseline CNN model with different dense layers. Evaluation criteria (from left to right) are Acc, Sn, and Sp, where \(z_{2}\) represents going through two dense layers: from dropping out 20% data, to 128 neurons with ReLU activation function, and to decision nodes with Sigmoid activation function; \(z_{3}\) represents going through one dense layer: from dropping out 50% data to decision nodes with Sigmoid activation function; \(z_{2}+z_{3}\) represents the combination of \(z_{2}\) and \(z_{3}\)Back to article page