Ask for some code detail

[ddmm2020]

Whether your code var_update_op = state_ops.assign(var, var_update) change the w1,i of COCOB-Backprop Algorithm?

In your paper Training Deep Networks without Learning Rates
Through Coin Betting
You said that:
Note that the update in line 10 is carefully defined: The algorithm does not use the previous wt,i in
the update. Indeed, this algorithm belongs to the family of the Dual Averaging algorithms, where the
iterate is a function of the average of the past gradients [Nesterov, 2009].

You do so in Algorithm 2 COCOB-Backprop line 10