Portfolio Selection Problem is amongst the most studied 
topics in economics and finance: in its basic formulation 
is concerned with finding the portfolio minimising the 
risk, given a minimum required level of returns. 
The basic model is formulated in the seminal work by 
Markowitz, in which the formulation is given by minimising 
the variance (as a risk measure) for a given level of 
return. In this formulation asset weights sum up to one 
and the problem is solvable with exact methods, but when 
adding additional features, it becomes untractable even for
small instances. So heuristic approaches have been 
exploited to solve realistic instance of portfolio selection.
In our approach, we devised a master-slave decomposition 
in which a local search is introduced to determine assets 
to be included in the portfolio, and a Quadratic Programming 
solver  "decides" the optimal assets' weights. Results show 
that this approach is effective in tackling different 
portfolio selection formulations.