Skip to main content

Course Roadmap

Quick overview​

  1. Foundations: What EC is, optimization basics, representation and operators.
  2. Genetic Algorithms: Holland's GA through real-coded, adaptive, and parallel variants.
  3. Evolution Strategies: (1+1)-ES through CMA-ES and natural evolution strategies.
  4. Other EC Paradigms: Genetic programming, differential evolution, EDAs, coevolution.
  5. Swarm Intelligence: PSO, ant colony optimization, and the nature-inspired zoo.
  6. Multi-Objective & Diversity: Pareto optimization, many-objective, quality-diversity.
  7. Neuroevolution & Deep Learning: NEAT, EC for NN training, ES for RL.
  8. Advanced Topics: Constraints, surrogates, hyper-heuristics.
  9. Rigor & Capstone: Benchmarking methodology, theory, and your capstone project.

Prerequisites​

  • Math: Linear algebra (eigenvalues, covariance matrices), multivariate calculus (gradients, Hessians), probability and statistics (Gaussian distributions, expectations, KL divergence), basics of optimization (gradient descent, convexity).
  • Coding: Python (NumPy, Matplotlib, basic OOP). Some PyTorch experience helpful for later modules.
  • Tools: Git, venv/uv, Jupyter or VS Code.

Books​

  • Introduction to Evolutionary Computing by Eiben & Smith - the standard modern textbook.
  • Evolutionary Computation: A Unified Approach by Kenneth De Jong - authoritative theoretical treatment.
  • Handbook of Evolutionary Computation by Back, Fogel & Michalewicz - comprehensive reference.
  • Adaptation in Natural and Artificial Systems by John Holland - the founding GA text.
  • The CMA Evolution Strategy: A Tutorial by Nikolaus Hansen - essential CMA-ES reading.
  • Ant Colony Optimization by Dorigo & Stutzle - definitive ACO reference.
  • Genetic Programming by Koza - the four-volume series that started GP.
  • Computational Intelligence: An Introduction by Engelbrecht - broader CI context.
  • Essentials of Metaheuristics by Sean Luke - accessible and free online.

Key papers​

  • Holland (1975): Adaptation in Natural and Artificial Systems - the GA foundation
  • Rechenberg (1973): Evolutionsstrategie - the ES foundation
  • Hansen & Ostermeier (2001): "Completely Derandomized Self-Adaptation in Evolution Strategies"
  • Storn & Price (1997): "Differential Evolution - A Simple and Efficient Heuristic for Global Optimization"
  • Kennedy & Eberhart (1995): "Particle Swarm Optimization"
  • Dorigo et al. (1996): "Ant System: Optimization by a Colony of Cooperating Agents"
  • Koza (1992): "Genetic Programming: On the Programming of Computers by Means of Natural Selection"
  • Deb et al. (2002): "A Fast and Elitist Multiobjective Genetic Algorithm: NSGA-II"
  • Stanley & Miikkulainen (2002): "Evolving Neural Networks through Augmenting Topologies" (NEAT)
  • Salimans et al. (2017): "Evolution Strategies as a Scalable Alternative to Reinforcement Learning"
  • Mouret & Clune (2015): "Illuminating search spaces by mapping elites" (MAP-Elites)
  • Wierstra et al. (2014): "Natural Evolution Strategies"

Key libraries​

  • cma (pycma) - reference CMA-ES implementation
  • nevergrad - Meta's gradient-free optimization platform
  • deap - Distributed Evolutionary Algorithms in Python
  • pyswarms - Particle swarm optimization
  • gplearn - Genetic programming / symbolic regression
  • pymoo - multi-objective optimization framework
  • neat-python - NEAT implementation
  • evosax - JAX-based ES library (GPU-accelerated)
  • qdax - Quality-Diversity algorithms in JAX
  • ioh - IOHprofiler benchmarking framework

Suggested schedule (flexible)​

TimeframeModulesFocus
Month 100-02Foundations, optimization basics, operators
Month 203-05Genetic algorithms, (1+1)-ES
Month 306-08Population ES, CMA-ES, NES
Month 409-12GP, DE, EDAs, coevolution
Month 513-15Swarm intelligence
Month 616-18Multi-objective, quality-diversity
Month 719-21Neuroevolution, EC for NNs, ES for RL
Month 822-24Constraints, surrogates, hyper-heuristics
Month 925-26Benchmarking, theory, capstone

For research-level mastery, expect 12-15 months including reading papers and contributing to open-source EC libraries.

Practical project ideas​

  1. EC Algorithm Showdown: Compare GA, CMA-ES, DE, and PSO on BBOB functions with statistical rigor.
  2. Evolve a robot controller: Use ES or NEAT to train a neural network policy for a locomotion task.
  3. Symbolic regression with GP: Rediscover physical laws from data using genetic programming.
  4. Quality-Diversity robotics: MAP-Elites for diverse robot gaits with damage adaptation.
  5. Multi-objective engineering: Solve a real trade-off problem with NSGA-II and visualize the Pareto front.
  6. ACO for logistics: Solve a real routing/scheduling problem with ant colony optimization.
  7. Surrogate-assisted optimization: Apply surrogate-assisted CMA-ES to an expensive simulation.
  8. Neuroevolution for game AI: Evolve a NEAT agent to play a game, visualize the topologies.
  9. Coevolution game strategies: Evolve competing strategies for Iterated Prisoner's Dilemma.
  10. DE for engineering design: Optimize a real engineering problem (antenna, airfoil, circuit).

Metrics of progress​

  • You can explain the EC landscape and when to use each algorithm family.
  • You can implement a GA, (1+1)-ES, CMA-ES, DE, and PSO from scratch.
  • You can use pycma, nevergrad, deap, and pyswarms to solve practical problems.
  • You can build a GP system for symbolic regression.
  • You can set up rigorous BBOB benchmarks and produce ECDF plots.
  • You can evolve a neural network controller for a locomotion task.
  • You can solve multi-objective problems and visualize Pareto fronts.
  • You can explain the No Free Lunch theorem and its practical implications.
  • You can read a new EC paper and reproduce its experiments.
  • You can critically evaluate a "novel nature-inspired algorithm" paper.

Was this page helpful?