Synchronous Parallel SGD $(12\mathrm{.}5$ points$)$ Now assume we use a synchronous ap$-$

proach where we wait for all the cores to finish calculating a gradient, and then we

aggregate the gradients and update the shared memory $($no gradient calculation happens

during aggregation or memory update$).$ If calculating the gradient on core $1,$ core $2,$ and

core $3$ takes $0\mathrm{.}4$s$,0\mathrm{.}45$s$,$ and $0\mathrm{.}5$s$,$ respectively, aggregating the gradients takes $0\mathrm{.}3$s$,$ and

writing an update in the memory takes $0\mathrm{.}25$s$,$ how long does it take for the cores to finish

$1$ parameter update?