CanSum Dynamic Programming C# memoization is still slow

I am trying to learn dynamic programming and I am following a 5h tutorial from freecodecamp but he is using js, and obj in js is easy with indexing, in C# I am trying to use dictionary
and I have to use .ContainsKey twice which I think is expensive on the performance.

public bool CanSum(int n, int[] numbers, Dictionary<int, bool> memo)
{
    if (n == 0) return true;
    if (n < 0) return false;

    if (memo.ContainsKey(n))
        if (memo[n]) return true;

    for (var i = 0; i < numbers.Length; i++)
    {
        var sum = n - numbers[i];

        if (CanSum(sum, numbers, memo))
        {
            if (!memo.ContainsKey(sum))
                memo.Add(sum, true);
            return true;
        }
    }

    return false;
}

How can I improve this? Is there a possibility to use an array?

Edit: Ok I did learn something new, I can new Dictionary as an empty object, like this new() { }

And code is like this:

if (memo.ContainsKey(n))           
     return memo[n];

if (n == 0) return true;
if (n < 0) return false;

for (var i = 0; i < numbers.Length; i++)
{
    var sum = n - numbers[i];

    if (CanSum(sum, numbers, memo))
    {
        memo[n] = true;
        return true;
    }
}

memo[n] = false;

return false;

First of all, the golden rule of optimization is to measure. So if you say things like "think is expensive" you should first learn how to actually check your assumptions. Because even experienced programmers are often wrong when they assume things about performance. There are several possible tools for doing measurements, Stopwatch is the simplest and most basic one, a performance profiler would give you the most data, and Benchmark .Net should be the most accurate one.

> Is there a possibility to use an array?

If the range of numbers are limited you can use an array. If n can be any int you would need a large array. Arrays should in general be faster than dictionaries, but se the first paragraph about assumptions. I'm guessing elements in your array need to have three states, true, false and unknown, aka trilean or ternary.

I would also be a bit wary about toy examples like this. First of all because I do not think they represent common problems very well. Second because techniques like "don't redo expensive calculations" should be applicable to many problems. And third, things like dictionary vs array is a type of micro-optimization (i.e. both have constant access time), but this is a large and complex topic in itself, and I would argue, more generally applicable than "dynamic programming".