Running Differential Evolution (DE)

This guide focuses on setting up and running Differential Evolution (DE) experiments using EvOC.

What is Differential Evolution (DE)?

Differential Evolution (DE) is a powerful population-based optimization algorithm, particularly effective for continuous optimization problems. It creates new candidate solutions by combining existing ones in a unique way, involving vector differences. In EvOC, DE is treated as a specific strategy available within the general "Evolutionary Algorithms (EA)" configuration section.

Configuring and Running DE

Follow these steps to set up and execute a DE algorithm:

1. Select Main Algorithm Type:

   • From the main EvOC Dashboard, click on the Evolutionary Algorithms (EA) option. DE is configured within this general EA framework.

2. Choose the de Strategy:

   • This is the crucial step for DE. Within the configuration screen, specifically select de from the available "Algorithm Strategy" options.

   

   Key Step

   Selecting the de strategy is essential to activate the Differential Evolution specific logic and parameters.

3. Configure DE Parameters:

   • Once de is selected, configure the relevant parameters. While some overlap with standard GAs (like Population Size, Generations, Evaluation Function), DE has specific parameters you'll need to set:
     • CR (Crossover Rate): Probability of incorporating parts of the mutant vector into the target vector.
     • F (Differential Weight): Scaling factor for the difference vectors, controlling amplification.
     • Individual Representation: Typically floatingPoint for DE.
     • Population Size and Number of Generations.
     • Evaluation Function.

   Parameter Details

   Consult the tooltips within the EvOC interface for specific guidance on recommended ranges or typical values for CR and F, as they can significantly impact performance.

4. Execute the Algorithm:

   • After defining your DE parameters, click the Execute Algorithm button.

Understanding and Using Your DE Results

The results interface for DE runs is similar to other traditional EAs:

Fitness & Best Solution

• Analyze the Fitness Plot showing the progression of minimum, average, and maximum fitness over the generations.
• Identify the Best Individual Fitness and the corresponding Best Individual (vector of values) found by the algorithm.

Sharing Your Run

• Use the Share Run button to share the DE configuration and results with other EvOC users via email (comma-separate multiple addresses).

  

Viewing and Downloading Logs

• Click Show Logs to inspect generation-by-generation fitness statistics.

• Click Download Logs within the log view to save the data (.txt) for external analysis.

  

Viewing the Generated Code

• Use Show Code to view the DEAP Python code implementing the DE logic based on your settings.

• Remember the Ask EvOC AI to Explain feature is available for code assistance.

  

Accessing Run History

• Find all your past runs, including DE experiments, in the View Previous Runs or View All Runs section of EvOC.

---

Next Steps:

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