What does this TypeError("invalid NaN comparison") mean? And how do I fix it?

New to Julia
, , , ,
1

I’m trying to use the function “solve” from sympy to solve for the values that satisfies a complicated equation. I am receiving the following error message: TypeError(“invalid NaN comparison”)

Does anyone know how to resolve this issue or understand where it is coming from? Any advice would be much appreciated!

Thanks

image001
image0011081×569 95.9 KB

2

hi, can you post what are term2 and x_hat?

3

x_hat = [symbols(“x_hat$i”,real=true) for i=1:n].reshape(n,1)
term2 = find_max(find_min(Q*x_hat,d),zeros(n,1))
n is an integer, Q is a nxn matrix of floats, d is also an nx1 vector

  1. term2 is in vague terms. In the real code it’s actually more complicated :frowning: But, hopefully this gets the main point across
  2. find_max and find_min are functions I made myself - the continuous form of minimum and maximum functions. (the reason why I wrote my own vs. the built-in max and min functions is because I originally wrote this in matlab and in matlab, it couldn’t take the minimum or maximum of symbolic terms)
  3. also in matlab, I used the function vpasolve() and it worked there. I am translating my matlab code into Julia to speed the process up

reference for find_min and find_max functions:

function find_min(A,B)
val = A - (abs.(A-B)+A-B)./2;
return val
end

function find_max(A,B)
val = B + (abs.(A-B)+A-B)./2;
return val
end

1 Like
4

hmm, got the same error. this is the full code to reproduce that error

import Sympy
n = 10

Q = rand(n,n)
d = rand(n,1)

function find_min(A,B)
  return  A - (abs.(A-B)+A-B)./2;
end

function find_max(A,B)
  return  B + (abs.(A-B)+A-B)./2;
end

x_hat = [Sympy.symbols("x_hat$i",real=true) for i=1:n].reshape(n,1)}
term2 = find_max(find_min(Q*x_hat,d),zeros(n,1))
Sympy.solve(term2-x_hat,x_hat)
5

@Grace_Chuan i think that the error lies in the find_min and find_max errors. you are creating an expression with Abs(sym) on term2-x_hat. changing those functions to:


function find_min(A,B)
    res = zeros(eltype(A),size(A)) #create an Array of the same size and element type
    for i in eachindex(A)
        if A[i] < B[i] #compare an store the correct branch instead of the abs function
            res[i] = A[i]
        else
            res[i] = B[i]
        end
    end
    return res
  end
  
  function find_max(A,B)
    res = zeros(eltype(A),size(A))
    for i in eachindex(A)
        if A[i] > B[i]
            res[i] = A[i]
        else
            res[i] = B[i]
        end
    end
    return res
 end

results in:

julia> Sympy.solve(term2-x_hat,x_hat)
Dict{Any, Any} with 10 entries:
  x_hat6  => 0.926800682408201 
  x_hat10 => 0.962863463691171 
  x_hat7  => 0.497643557077236 
  x_hat3  => 0.211146695704730 
  x_hat2  => 0.568369033677095 
  x_hat5  => 0.187652949522494 
  x_hat4  => 0.392861232103788 
  x_hat9  => 0.559299830282101 
  x_hat1  => 0.431629987943204 
  x_hat8  => 0.320110979997330

of course, there is the corner case of A[i] == B[i] but that is outside the scope of this problem, and more of the specific domain that are you working on

1 Like