Binary Variable

Specific Domains Optimization (Mathematical)
1

I got in my problem some binary variable, thats a decision to install or not an equipment in a house. Among a lot of other variables and constraints, my problem is something like this:

N = 5; # number of houses
@variable(m, a[i=1:N], Bin)
@objective(m, Min, sum(a[i] for i = 1:N))

But for a larger N (around 30+), the problem gets to big to be solved in an acceptable time (using a Branch-and-Bound solver).
So, to reduce the number of binary variables, i was thinking in just choose some a[i] to be binary. For example, binary for i = 2:3:

@variable(m, a[i=2:3], Bin)

But i’m getting this error:

ERROR: LoadError: Failed attempt to index JuMPArray along dimension 1: 1 ∉ 2:3

Does anyone knows any trick in how work around this?

2

1 ∉ 2:3 says there was no index 1 in 2:3(i think…), it might’ve come from @objective if it wasn’t updated to 2:3 as well.
Otherwise it could just need to start from 1

…unless it really means dimensions when it says #1 is missing, but it doesn’t seem too likely

3

If you can to create 3 variables but only the last two are binary you need to do

julia> using JuMP
julia> model = Model()
julia> @variable(model, x[i=1:3], category = (i in 2:3 ? :Bin : :Cont))
julia> print(model)
Min 0
Subject to
 x[1]
 x[2] ∈ {0,1}
 x[3] ∈ {0,1}

Note that the above will work only with JuMP up to JuMP v0.18. With JuMP master, the syntax is now

@variable(model, x[i=1:3], binary = i in 2:3)
1 Like
4

Yeah, this was the problem… i forgot to change the range of i in the @objective and in some @constraint.

Thanks!

1 Like
5

btw There seems to be something wrong with the JuMP documentation page, i had to Ctrl-F5 and then Esc to catch the page text before it disappeared
/Firefox 60.0.1 on windows 7