[Retracted] Classroom Behaviour of Talent Cultivation in Colleges and Universities Supported by Deep Learning Technology

Algorithm 1

Convolution neural network with an attention mechanism.

Phase 1: traditional phase

Set all of the CNN-AM’s weights and biases to a low number.

Set the learning rate so that

repeat

for to , do

disseminate the pattern over the network, propagate the pattern throughout the network

for to the number of neurons in the output layer

Error detection

end for

for layers to 1, do

for maps to , do

find a back-propagated error factor

end for

end for

for , I do

for to , do

for all weights of the map, a do

Find

Weights and biases should be updated

end for

end for

Calculate the Mean Square Error (MSE1)

Until or

Phase 2: knowledge transfer repeat

from to PS (number of new training samples) propagate the pattern across the network

for to the number of neurons in the last convolutional layer ()

find the output of the last layer of the convolutional layer.

 = (,, …….. )

Find using the TSL framework (Section 3)

end for

Phase 3: update your weight for the transfer learning phase.

for to PR

Train the feedforward layers (layers after the last convolutional layer) using available in phase

Gradient (second) descend algorithms [14] are a viable option.

end for

Find MSE2

Until or