Question about the Loss term "L_AReg"

1. Thanks for your excellent work. I have question that it seems the computation way of "L_AReg" in the paper is different from that in the code,  in the paper, L_AReg = r1 + r2 + ||r1-r2||_2 while in the codes, it changes to r1+r2 or r1-r2, Could you please explain this issue? Very glad to receive  your reply.