Research Paper Volume 16, Issue 10 pp 9072—9105

Figure 8. The neural network model determines the function f as shown in Equation 3, which is divided into an input layer (m neurons, m = 16782), a hidden layer (3 neurons), and an output layer (2 neurons). Where each neuron corresponds to a gene expression in a certain sample, thus a total of m genes corresponds to m neurons. Sigmoid function as an activation function in hidden layers. A Softmax layer is added to the output layer to transform the output of output layer to probability. Therefore, the output of function f represents the probability of having AD or not.