Data-Driven Quality Prediction of Batch Processes Based on Minimal-Redundancy-Maximal-Relevance Integrated Convolutional Neural Network

Table 1

Specific structure of fixed mRMR-CNN.

Type	Kernel size/stride	Output size	Parameters	Trainable
Convolution	3 × 3/1	50 × 16 × 10	100	True
Max pooling	2 × 2/2	25 × 8 × 10	—	—
Convolution	3 × 3/1	25 × 8 × 20	1820	True
Max pooling	2 × 2/2	12 × 4 × 20	—	—
Feature selection	—	336	336	False
Dropout (50%)	—	336	—	—
Full-connection	—	40	13480	True
Dropout (50%)	—	40	—	—
Output	—	1	41	True