The plots are not for d=0. That confused me.
Robust regression is obtained by using a l1 loss in place of least squares. Removes outliers. For d=0, computes the median instead of the mean. https://en.wikipedia.org/wiki/Robust_regression …pic.twitter.com/A77Rt3imkZ
-
-
-
Plots are for d=10. d=0 would have produced an horizontal line.
- Još 2 druga odgovora
Novi razgovor -
-
-
Sorry for knitpicking, but isnt it more ‚augmenting‘ least squares loss with additional l1 term, rather than ‚replacing‘ it?
-
Are you thinking about a Lasso? If so, no, this is about changing the loss function from l2 to l1. It is the residual error which is sparse (to cope with outliers) rather than the coefficients (to perform model selection). But you can use the same algorithms.
- Još 1 odgovor
Novi razgovor -
-
-
Why d=0?
-
It is for constant polynomial
- Još 1 odgovor
Novi razgovor -
-
-
@EveryWhereChem an example of using MAE as the loss function instead of MSE - the effect is to discount outliers (for good or ill)!Hvala. Twitter će to iskoristiti za poboljšanje vaše vremenske crte. PoništiPoništi
-
-
-
Can’t believe I’ve never heard of this whole branch of “robust statistics” before! Amazing.
Hvala. Twitter će to iskoristiti za poboljšanje vaše vremenske crte. PoništiPoništi
-
-
-
How do you generate your excellent visualizations?
Hvala. Twitter će to iskoristiti za poboljšanje vaše vremenske crte. PoništiPoništi
-
-
-
This is a great pic! Thanks for sharing! Do you know which software produces these animations?
Hvala. Twitter će to iskoristiti za poboljšanje vaše vremenske crte. PoništiPoništi
-
Čini se da učitavanje traje već neko vrijeme.
Twitter je možda preopterećen ili ima kratkotrajnih poteškoća u radu. Pokušajte ponovno ili potražite dodatne informacije u odjeljku Status Twittera.