The high-cardinality of mean-variance portfolios is a concern in practice because it increases transaction costs and management fees. Therefore, we propose a method to resolve the cardinality problem by applying the semi-definite relaxation method to a cardinality constrained optimal tangent portfolio selection model. We find that the relaxed model becomes a semi-definite programming problem that is efficiently solved with existing optimization solvers. Numerical analyses with historical stock returns confirm that the proposed relaxed model effectively constructs sparse tangent portfolios. (C) 2016 Elsevier B.V. All rights reserved