m and n are greater than 0,this has a convex function.The goal is to make it less than a number. How do you program it in cvx?

How to avoid situations like {convex} .^ {2}

`square_pos(inv_pos(m)+inv_pos(n)) <= number`

But it’s probably better to use

`inv_pos(m)+inv_pos(n) <= sqrt(number)`

Thank you for your generous answer! The code seems to be correct, and the constraints of the square_pos（） are satisfied, but there seems to be something wrong with the output of the function.

cvx_begin

```
variables m n t x positive
minimize(m)
subject to
square_pos(inv_pos(m)+inv_pos(n)) <= 50;
```

cvx_end

inv_pos(m)+inv_pos(n)

square_pos(inv_pos(m)+inv_pos(n))

output:

Status: Solved

Optimal value (cvx_optval): +0.141421

ans =

```
7.0711
```

ans =

```
1
```

Squared with^2 is correct ,Is this normal?

I don’t know why your `square_pos(inv_pos(m)+inv_pos(n))`

comes out to 1. Mine comes out to 49.999987213706319, which is within solver tolerance of 50.

Note that the solve will set the optimal value of `n`

to a very large number, so that `1/n`

is very small. Also note that `variables m n t x positive`

did not declare `m`

,`n`

, and `t`

as positive, but actually declared `positive`

as a variable. However, use of `inv_pos`

has the effect of constraining `m`

and `n`

to be positive. If you want to declare a variable as positive, you can only declare a single variable as such per statement, using `variable`

, not `variables`

. That is so that the attribute, such as `positive`

will not be confused as being a variable name in a `variables`

statement.

There are actually 2 different `square_pos`

functions. One of them is used on CVX expressikons, and is what is used during the optimization. The other is what is applied to double precision arguments, and is what is called after cvx_end, when `m`

and `n`

are double precision values which CVX set to their optimal values. So perhaps something is corrupted with or another version of `square_pos`

was called in your case. Do you have another `square_pos`

from something? I am not aware of any `square_pos`

, other than from CVX, being in any publicly available MATLAB toolbox or add-on, or base MATLAB.