$8.$ Recall that in backpropagation, for each network weight, weights are updated by

$\backslash$geotriangle $$ ji $=$ j $$ ji $+$ momentum$-$term

where w ji is the weight from unit i to unit j$,$ xji is the input coming from unit i to unit j$,$ and $$j is the

error term at unit j$.$

$(8$a$)$ Suppose you are training a multilayer neural network. You are about to update weight w ji $.$

Suppose you have the following values: Current value for weight w ji $=0\mathrm{.}1$ xji $=1$

$!=0\mathrm{.}1$

Previous value of $\backslash$geotriangle $$ji $=0\mathrm{.}2$

Learning rate $\backslash$eta $=0\mathrm{.}1$ Momentum

parameter $\backslash$alpha $=0\mathrm{.}2$ What is the new value

of w ji $?$

$(8$b$)$ In one sentence, what is the purpose of the momentum term?