Solving MCP with PathSolver with indexing

Makes sense! Thanks.

From your experience, does setting a floor to denom_i[s], i.e.,

denom_i .= A * collect(x[i:J:end])
            for s in 1:S
                v[s] = A[s, j] / maximum(denom_i[s],epsilon)^2
            end

Is a good idea? Or will this mostly confuse the solver?