Automated Classification and Segmentation in Colorectal Images Based on Self-Paced Transfer Network

Algorithm 1.

Flowchart of STVGG algorithm.

Input: Training set , represents the training data, and is the data label.

Initialization parameter: “age parameter”, a suitable initial value is given according to the approximate value range of the presample training error value; initialize the sample weight vector .

Model training parameter settings: total number of training iterations epoch, minimum batch size for training and verification, initial learning rate during model training , and decay rate of learning rate ; update increment of age parameters .

a) Calculate network weights and bias by Eq. (2)

b) Calculate and update loss function

c) Calculate self-paced regular term and update weight vector

d) Calculate and update

e) Update age parameters and learning rate , ,

f) Repeat steps a to e until the number of iterations

output: network weights and bias

Algorithm 1.