A genetic system based on simulated crossover: stability analysis and relationships with neural nets

Area of Science:

• Computational Biology
• Genetics
• Artificial Intelligence

Background:

• Marginal distribution genetic models analyze gene sequence evolution.
• Neural networks offer computational frameworks for complex systems.
• Understanding the relationship between genetic dynamics and neural computation is an emerging area.

Purpose of the Study:

• To establish a connection between marginal distribution genetic models and neural networks.
• To explore the behavior of finite population genetic systems in relation to infinite population models.
• To characterize attractors in genetic systems using neural network concepts.

Main Methods:

• Developing a marginal distribution genetic model incorporating gene crossover.
• Analyzing the asymptotic behavior of finite population genetic systems.
• Mapping genetic system attractors to equilibrium points of a modified Hopfield network.
• Utilizing fitness as a Lyapunov function for the neural network model.

Main Results:

• A lower bound on population size is established for approximating finite systems with infinite ones.
• Assumptions on fitness and chromosomes ensure consistency between finite and infinite genetic systems over long trajectories.
• Attractors of the infinite population genetic system correspond to equilibrium points in a discrete neural network variant.
• Fitness functions serve as Lyapunov functions for this discrete neural network.

Conclusions:

• The relationship between marginal distribution genetic systems and neural networks is more general than previously demonstrated.
• This work provides a novel framework for understanding genetic evolution through the lens of neural computation.
• The findings suggest potential applications in computational neuroscience and evolutionary algorithms.