DarwinGA 1.0.0

🧩 Creating a Custom Evolutionary Problem

DarwinGA is designed to be problem-agnostic.

That means you don’t solve a problem directly — you define how a solution behaves, and the algorithm evolves it.

To use DarwinGA, you only need to define 3 things:

1️⃣ Define your Solution (Chromosome)

This represents a candidate solution.

public class MySolution
{
    public double Price { get; set; }
}

👉 This is what evolves.

You can model anything:

• numbers
• arrays
• objects
• strategies

2️⃣ Define the Fitness Function

This is the most important part.

It tells the algorithm:

“How good is this solution?”

public class MyFitnessEvaluator : IFitnessEvaluator<MySolution>
{
    public double Evaluate(MySolution solution)
    {
        // Example: maximize profit
        double demand = 100 - solution.Price;
        double profit = solution.Price * demand;

        return profit;
    }
}

👉 The algorithm will try to maximize this value.

3️⃣ (Optional) Customize Mutation & Crossover

You can control how solutions evolve.

Mutation

public class MyMutation : IMutationStrategy<MySolution>
{
    public void Mutate(MySolution solution)
    {
        solution.Price += Random.Shared.NextDouble() * 2 - 1;
    }
}

Crossover

public class MyCrossover : ICrossoverStrategy<MySolution>
{
    public MySolution Crossover(MySolution a, MySolution b)
    {
        return new MySolution
        {
            Price = (a.Price + b.Price) / 2
        };
    }
}

👉 These define how evolution behaves.

🚀 Running the Algorithm

Once defined, you plug everything into the engine:

var ga = new GeneticAlgorithm<MySolution>(
    populationSize: 100,
    mutationRate: 0.05,
    crossoverRate: 0.7,
    fitnessEvaluator: new MyFitnessEvaluator(),
    mutationStrategy: new MyMutation(),
    crossoverStrategy: new MyCrossover()
);

ga.Initialize(() => new MySolution
{
    Price = Random.Shared.NextDouble() * 100
});

ga.Run(generations: 200);

var best = ga.GetBestSolution();

Console.WriteLine($"Best price found: {best.Price}");

🧠 How to Think in Evolutionary Terms

Instead of asking:

❌ “How do I solve this problem?”

You ask:

✅ “What does a solution look like?” ✅ “How do I measure if it's good?” ✅ “How can I slightly modify it?”

That’s it.

The algorithm does the rest.

💡 Key Insight

A good evolutionary setup depends on:

• ✔️ A meaningful fitness function
• ✔️ A representation that can evolve smoothly
• ✔️ Balanced mutation (not too random, not too rigid)

⚠️ Common Mistakes

• ❌ Fitness function too simple → no evolution
• ❌ Mutation too aggressive → chaos
• ❌ No diversity → early stagnation
• ❌ Overfitting to a specific case

🔥 Real Example Ideas

You can build custom evolutions for:

• 💰 Pricing optimization
• 📦 Warehouse distribution
• 🚚 Delivery routes
• 🧠 AI parameter tuning
• 🎯 Strategy optimization

🧬 Final Mental Model

Think of DarwinGA as:

A system where you define the rules of survival, and solutions fight to become the best.