Backpropagation based on TensorFlow

This paper mainly introduces the methods and processes of backpropagation. First, the principle of a single neuron is introduced. Emphasis is placed on the concepts of weights, biases, gradients in calculations, and the chain rule. The final result is achieved through programming on a computer with a specific case. Secondly, the working principle of forward propagation is introduced, and a diagram of a three - layer neural network is output through hands - on operation on a computer. Finally, the working principle and calculation process of backpropagation are emphasized. The backpropagation case is implemented on a computer with an actual case, and the correct result is output.