I think it could be a good way-around using the average pooling built-in function and then multipliying by the size of the pooling kernel ('Stride').
Here's an example having X as 1D array and applying an average pooling with N-cells stride.
N=4;
X= [1,2,3,4,5,6,7,8,9,10,11,12]
X_dl= dlarray(X,'S');
Y_dl= avgpool(X_dl,N,'Stride',N);
Y_avg_pooling= extractdata(Y_dl);
Y_sum_pooling= (Y_avg_pooling*N)'
