add second order cone type constraint

Specific Domains Optimization (Mathematical)
1

Hi,

I am new to Julia and JuMP, and I am learning to solve SOCP with integer constraint.

The only example I can find online shows I should add the constraint in the following way.

@constraint(m, norm(2x[i] - i for i=1:n if c[i] == 1) <= 1)

Is there a way to call norm function with out a looping of iterator with in it ? For example, in my problem, a segment of code would be

for t in 1:n 
    for i in 0:t
        for j in ( t + 1 ):( n + 1 )
            # add socp constraint here
            @constraint( model , dd[ k , i , j , t , 1 ] == sqrt( generationCostA( k , t ) )  * x[ k , i , j , t ] )
            @constraint( model , dd[ k , i , j , t , 2 ] == 0.5 * ( d[ k , i , j , t ] - y[ k , i , j ] ) )
            @constraint( model , norm( dd[ k , i , j , t , kk ] for kk in 1:2 ) <=  0.5 * ( y[ k , i , j ] + d[ k , i , j , t ] )  )
        end
    end
end

the vector within norm function has dimension 2. I tried to use norm([ x , y ]) , but it does not work. So I have to create dummy variables dd , and then I can use iterator within norm function.

Can I do it without introduce the dummy variables? Thanks for your time of reading, and much appreciated if you can help me on this!

Qm

1 Like
2

Welcome :slight_smile:

What doesn’t work? Maybe this should’ve been norm([x ; y])? It’s hard to say more without seeing code that reproduces the issue.

3

Hi,

Thanks for your reply!

When I add constraint in the way of calling norm( [ x , y ] ) like the following

@constraint( model , norm( [ sqrt( generationCostA( k , t ) )  * x[ k , i , j , t ]  , 0.5 * ( d[ k , i , j , t ] - y[ k , i , j ] ) ] ) <=  0.5 * ( y[ k , i , j ] + d[ k , i , j , t ] )  )

the error says norm not defined

ERROR: LoadError: UndefVarError: norm not defined
Stacktrace:
 [1] macro expansion at C:\Users\tommyricardo\.julia\packages\JuMP\PbnIJ\src\macros.jl:492 [inlined]
 [2] macro expansion at C:\Users\tommyricardo\.julia\packages\JuMP\PbnIJ\src\macros.jl:493 [inlined]
 [3] top-level scope at F:\a__top_secret\multiple_units_quadratic.jl:166 [inlined]
 [4] top-level scope at .\none:0
in expression starting at F:\a__top_secret\multiple_units_quadratic.jl:161

same thing happens if I called norm( [ x ; y ] ) norm( x , y ) or norm( x ; y )

@constraint( model , norm( [ sqrt( generationCostA( k , t ) )  * x[ k , i , j , t ]  ; 0.5 * ( d[ k , i , j , t ] - y[ k , i , j ] ) ] ) <=  0.5 * ( y[ k , i , j ] + d[ k , i , j , t ] )  )

ERROR: LoadError: UndefVarError: norm not defined
Stacktrace:
 [1] macro expansion at C:\Users\tommyricardo\.julia\packages\JuMP\PbnIJ\src\macros.jl:492 [inlined]
 [2] macro expansion at C:\Users\tommyricardo\.julia\packages\JuMP\PbnIJ\src\macros.jl:493 [inlined]
 [3] top-level scope at F:\a__top_secret\multiple_units_quadratic.jl:166 [inlined]
 [4] top-level scope at .\none:0
in expression starting at F:\a__top_secret\multiple_units_quadratic.jl:161

It only works if I defined new variables dd to make a copy of x,y and call norm( dd[ i ] for i in 1:2 ).

@constraint( model , dd[ k , i , j , t , 1 ] == sqrt( generationCostA( k , t ) )  * x[ k , i , j , t ] )
@constraint( model , dd[ k , i , j , t , 2 ] == 0.5 * ( d[ k , i , j , t ] - y[ k , i , j ] ) )
@constraint( model , norm( dd[ k , i , j , t , kk ] for kk in 1:2  ) <=  0.5 * ( y[ k , i , j ] + d[ k , i , j , t ] )  )

But in this way, I have to create a lot of dummy variables dd. So I am just wondering if there is a way to call norm for adding the constraint without doing a loop?

I did not put all the codes because it is a bit long. But I think this is the part that produce the error, if you need to see the other part as well, please let me know.

Thanks again for your help.

Qm

4

You need to do using LinearAlgebra for norm to be available.

5

Thanks! It works after I add LinearAlgebra packages.