[Retracted] CNN-LSTM Hybrid Model for Kinematic Feature Analysis and Parabolic Radian Prediction in Basketball Videos

Algorithm 1

RAdam-based optimization algorithm.

	Input: Define learning rate , initial parameters , numerical stability , gradient first-order decay coefficient , gradient second-order decay coefficient , and optimization objective function .
	Output: final parameters.
	Initialize simple moving average (SMA)
	While do
	Calculate the gradient at step t
	Calculate the second-order moment of the shift
	Calculate the correction for shift deviation
	Calculate the maximum value of SMA
	if then
	Calculate the correction value of the second-order moment of the moving volume ;
	Update of parameters using adaptive momentum
	Else
	Update of parameters using nonadaptive momentum
	Return