12:00 minutes , can you explain why that particular feature is important ? from the graph I can only make this assumption that they are not overlapping that's why but I want know how did you reach to that conclusion whether that certain feature is important or not ? Please someone explain this .
HI AMIT, your interrogation is indeed very interesting. This is an example that showcases why sometimes univariate feature selection is not relevant. If you look at that unique feature, the distributions of the feature are not overlapping and that should traditionally be indicative of a good separability. However, if you join that feature to another one and you look at both of them with your two-dimensional microscope, you start to realize it does not add much to the table. Also, this might happen if you add several additional features to them, even though this could be tricky to visualize. I believe this was the intuition behind this part; to look at features from a whole global point of view rather than a singular and doubtful one.
![[UA] Eternal Fire проти NAVI | EPL Season 20 Malta](http://i.ytimg.com/vi/chvlsDygrZM/mqdefault.jpg)
12:00 minutes , can you explain why that particular feature is important ? from the graph I can only make this assumption that they are not overlapping that's why but I want know how did you reach to that conclusion whether that certain feature is important or not ? Please someone explain this .
HI AMIT,
your interrogation is indeed very interesting. This is an example that showcases why sometimes univariate feature selection is not relevant.
If you look at that unique feature, the distributions of the feature are not overlapping and that should traditionally be indicative of a good separability. However, if you join that feature to another one and you look at both of them with your two-dimensional microscope, you start to realize it does not add much to the table. Also, this might happen if you add several additional features to them, even though this could be tricky to visualize.
I believe this was the intuition behind this part; to look at features from a whole global point of view rather than a singular and doubtful one.
Thanks for the lecture, but have to work on "yeah" part every 10 sec.
nope, it is ok