Running a Traditional Evolutionary Algorithm

This guide details how to set up and run a "traditional" Evolutionary Algorithm (EA) using EvOC.

What is a "Traditional EA" in EvOC?

In the context of EvOC, a "traditional EA" refers to evolutionary algorithms that do not involve Genetic Programming (GP). This typically includes approaches like standard Genetic Algorithms (GA) or Differential Evolution (DE) that operate on fixed-length representations (like lists of numbers or binary strings) rather than trees.

Configuring and Running the EA

Follow these steps to build and execute your algorithm:

1. Select Algorithm Type:

   • From the main EvOC Dashboard, click on the Evolutionary Algorithms (EA) option. This section houses the traditional EA configurations.

   

2. Choose Algorithm Strategy:

   • Select the specific EA strategy you want to use (e.g., eaSimple, eaMuPlusLambda, de).

   Tip

   Most strategies listed under the "Evolutionary Algorithms (EA)" section are suitable for traditional EA approaches. The primary exception is typically strategies explicitly designed for Genetic Programming (GP), which have their own section.

   Differential Evolution (DE) Note

   While de (Differential Evolution) is listed here, its specific parameter requirements might differ slightly from other generational GAs like eaSimple. Refer to parameter tooltips within EvOC for details.

3. Configure Parameters:

   • Work through the configuration options presented in the UI, setting parameters such as:
     • Optimization objective (Maximize/Minimize Weights)
     • Individual representation (e.g., binaryString, floatingPoint)
     • Population size
     • Number of generations
     • Mating (Crossover) function and probability
     • Mutation function and probability
     • Selection function
     • Evaluation function
   • The video below demonstrates the process of selecting options and filling in values for a sample configuration.

4. Execute the Algorithm:

   • Once you are satisfied with your configuration, scroll to the bottom and click the Execute Algorithm button. EvOC will start processing your experiment.

Understanding and Using Your Results

After the execution status changes to "Completed", the results interface will appear, providing several options:

Visualizations & Fitness

• Observe the Fitness Plot to see how the minimum, average, and maximum fitness evolved over generations.
• Check the Best Individual Fitness section for the final best solution found and its fitness value.

Sharing Your Run

• Click Share Run to share this specific experiment configuration and results with other registered EvOC users.

• Enter their email addresses (comma-separated for multiple users) in the pop-up window.

  

Viewing and Downloading Logs

• Click Show Logs to view detailed generation-wise statistics (e.g., average fitness, min/max, standard deviation).

• Click the Download Logs button within this view to save the log data as a .txt file for external analysis.

  

Viewing the Generated Code

• Click Show Code to inspect the Python code generated by EvOC based on your configuration. This code utilizes the DEAP library.

  

• You can also use the integrated Ask EvOC AI to Explain feature here if you need clarification on the code.

  

Accessing Run History

• To see all your past experiments, navigate back to the main dashboard sections and look for the View Previous Runs or View All Runs button/link (located near the bottom or accessible from the initial algorithm selection screen).

---

Next Steps:

• Explore configuring Genetic Programming (GP).
• Learn about Particle Swarm Optimization (PSO).
• See how to use EA for ML Tuning.