Ho ho ho, just take a note and support Xiaohong. If you don’t learn TTS in two months, you will not be able to learn it. The deep learning neural network algorithm is composed of forward propagation and backward propagation #1. First, the prediction result and loss are calculated through forward propagation and then through reverse propagation. Directional propagation calculates the partial derivatives of the loss function with respect to each function (w, b) and performs gradient descent on these parameters. Then a new round of forward propagation calculation is used, and repeated training makes the prediction more and more accurate. The feature vector activation function sigmoid(z) function is used to map data between 0 and 1. If the activation function is not used, the AI's intelligence will not improve. If there is no activation function applied to the trained function, the neural network can only calculate linear functions, so it cannot solve complex problems sigm