I am trying to figure out how to solve:
where rp=α⊺ω and U is a sphere centered at α with radius equal to χ|α| . ( ω is a vector or weights and χ lies between 0 and 1.)
i.e. we are trying to minimize rp inside the sphere centered at α.
The authors end up with the following solution:
minU rp = α⊺ω − χ |α| |ω|
The authors hint at Bayesian estimation but I am not familiar with it. Any ideas as to how they may have arrived at this would be great. Many thanks in advance.