BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Random Planted Forest: a directly interpretable tree ensemble - En
 no Mammen (Heidelberg University)
DTSTART:20220304T140000Z
DTEND:20220304T150000Z
UID:TALK168119@talks.cam.ac.uk
CONTACT:Qingyuan Zhao
DESCRIPTION:We introduce a novel interpretable and tree-based algorithm fo
 r prediction in a regression setting in which each tree in a classical ran
 dom forest is replaced by a family of planted trees that grow simultaneous
 ly.\nThe motivation for our algorithm is to estimate the unknown regressio
 n function from a functional ANOVA decomposition perspective\, where each 
 tree corresponds to a function within that decomposition.\nTherefore\, pla
 nted trees are limited in the number of interaction terms.  The maximal or
 der of approximation in the ANOVA decomposition can be specified or left u
 nlimited. If a first order approximation is chosen\, the result is an addi
 tive model. In the other extreme case\, if the order of approximation is n
 ot limited\, the resulting model places no restrictions on the form of the
  regression function. In a simulation study we find encouraging prediction
  and visualisation properties of our  random planted forest method.\nWe al
 so develop theory for an idealised version of random planted forests. In p
 articular\, for an additive model we show that the idealised version achie
 ves asymptotically optimal one-dimensional convergence rates of order $n^{
 -2/5}$ up to a logarithmic factor. The talk reports on joint work with Mun
 ir Hiabu (Copenhagen) and Joseph Theo Meyer (Heidelberg).
LOCATION:https://maths-cam-ac-uk.zoom.us/j/93998865836?pwd=VzVzN1VFQ0xjS3V
 DdlY0enBVckY5dz09
END:VEVENT
END:VCALENDAR