L-System vs. Matrix Neural Network Genetic Encoding Performance in a Novel World

A 2026 experiment by Alexander Stuy compares L-System (Lsys) and Direct Matrix genetic encodings for evolving Hebbian neural networks in an artificial world. The dataset includes results from 24 experimental runs, showing Lsys encoding achieved a mean maximum food count of 3802 ± 197 at generation 1000, a 2.74x advantage over Matrix encoding. The data supports findings on convergence speed, reliability, and generalization to a novel maze environment.

Use Cases

• Benchmarking genetic encoding methods for neural network evolution based on performance metrics like food count and coefficient of variation.
• Studying generalization capabilities of evolved networks based on transfer tests to a novel maze environment.
• Analyzing the impact of initial population structure on evolutionary performance based on the MatrixLSG control condition.
• Investigating the relationship between genetic alphabet compression and algorithm reliability based on the comparison of Lsys and Matrix encoding consistency.

Strengths

• Includes quantitative results from 24 experimental runs, providing a basis for statistical comparison.
• Reports specific performance metrics: Lsys encoding achieved a mean maximum food count of 3802 ± 197 at generation 1000.
• Documents a clear performance advantage: Lsys showed a 2.74x performance gain and an 8.5-fold improvement in consistency over Matrix encoding.
• Tests generalization with transfer to a novel maze environment, where Lsys achieved a 5.82x advantage.

Limitations

• Row count is unknown, which may limit suitability assessment.
• Column-level documentation is absent; field semantics must be inferred after download.
• The dataset is very small (22.8 KB), indicating limited scope, likely containing summary results rather than raw evolutionary traces.

Provenance

Source

Alexander Stuy via figshare

Collection Method

Generated from artificial life simulations comparing two neural network genetic encoding methods.

Freshness

Last updated 2026-05-03 08:19:11; freshness should be verified.