|
<frame title="knn">
|
|
<list>
|
|
<e>kNN algorithm can also be used for AD</e>
|
|
<e>if the k closest point is further away, a sample is considered more anomalous</e>
|
|
<e>$r=\frac{k}{2N\cdot pdf}$</e>
|
|
<e>Powerful method, as it can model the pdf directly</e>
|
|
</list>
|
|
</frame> |