API

The API of DEAP is built around a clean and lean core composed of two simple structures: a creator and Toolbox. The former allows creation of classes via inheritance and composition and the latter contains the tools that are required by the evolutionary algorithm. The core functionalities of DEAP are levered by several peripheral modules. The tools module provides a bunch of operator that are commonly used in evolutionary computation such as initialization, mutation, crossover and selection functions. It also contains all sort of tools to gather information about the evolution such as statistics on almost every thing, genealogy of the produced individuals, hall-of-fame of the best individuals seen, and checkpoints allowing to restart an evolution from its last state. The algorithms module contains five of the most frequently used algorithms in EC: generational, steady-state, , , and CMA-ES. These are readily usable for the most common population layouts. Specific genetic programming tools are provided in the gp module. A complete and efficient (and stand alone) parallelization module, base on MPI, providing some very useful parallel method is provided in the dtm module. Finally, common benchmark functions are readily implemented in the benchmarks.

• 1. Core Architecture
  • 1.1. Creator
  • 1.2. Toolbox
  • 1.3. Fitness
• 2. Evolutionary Tools
  • 2.1. Operators
  • 2.2. Statistics
  • 2.3. Hall-Of-Fame
  • 2.4. Evolution Logger
  • 2.5. Checkpoint
  • 2.6. History
• 3. Algorithms
  • 3.1. Complete Algorithms
  • 3.2. Variations
• 4. Covariance Matrix Adaptation Evolution Strategy
• 5. Genetic Programming
• 6. Distributed Task Manager Overview
  • 6.1. DTM Main Features
  • 6.2. Introductory example
  • 6.3. Functions documentation
  • 6.4. DTM launching
  • 6.5. Troubleshooting and Pitfalls
• 7. Benchmarks
  • 7.1. Continuous Optimization
  • 7.2. Binary Optimization
  • 7.3. Symbolic Regression
  • 7.4. Moving Peaks Benchmark
  • 7.5. Benchmarks tools