*to be continued and implemented as module....
The lagrangian multiplier tells that to optimize such
Then the optimization is later continued with the operations
The objective of the function is the optimized-margin (
, that is maximizing
It turns out that
- maximizing
$1/|w|$ is equal to minimizing$\|w\|$ - and, minimizing
$\|w\|$ is equal to minimizing$\|w\|^2$
also introduce the factor
- instead of minimizing
$\|w\|$ , minimize the$\frac{1}{2}|w|^2$
Why not maximizing 1/||w|| directly?
Because maximizing 1/|w| directly is complicated (it will takes several more steps
|
Overall, the constrained optimization defined as
subject to