“…While statistical properties of the convex regression estimator are well-established (Hildreth 1954, Hanson and Pledger 1976, Wu 1982, Fraser and Massam 1989, Mammen 1991, Groeneboom, Jongbloed, and Wellner 2001, Turlach 2005, Birke and Dette 2007, Chang, Chien, Hsiung, C.-C.Wen, and Wu 2007, Meyer 2008, Kuosmanen 2008, Shively, Walker, and Damien 2011, Seijo and Sen 2011, Lim and Glynn 2012, Lim 2014, the convex regression estimator suffers from computational inefficiency. Minimization of ψ n over C can be formulated as a quadratic program (QP) with (d + 1)n decision variables and n 2 constraints (Kuosmanen 2008).…”