Bardo Release Notes

Version 0.1.0

This is the initial release of Bardo, a focused neuroevolution library for Elixir.

Key Features

• Topology and Parameter Evolving Neural Networks (TWEANN): Bardo evolves both the structure and weights of neural networks over time.
• Efficient ETS-based Storage: Simple and fast in-memory storage with periodic backups.
• Modular Sensor/Actuator Framework: Easily connect networks to different environments.
• Built-in Evolutionary Algorithms: Includes selection algorithms and mutation operators.
• Substrate Encoding: Hypercube-based encoding for efficient pattern recognition.
• Example Environments:
  • XOR: Simple logical problem
  • Double Pole Balancing: Classic control benchmark
  • Flatland: Predator-prey simulation
  • FX: Basic forex trading simulation

Changes in This Release

• Initial public release with Apache License 2.0
• Removed experimental HTM (Hierarchical Temporal Memory) implementation
• Updated documentation for hex.pm publication
• Full test coverage across core functionality

Known Issues

• There are some compiler warnings that need to be addressed in future releases
• Some API functions need better documentation
• Certain modules implement behaviours incorrectly (especially in example code)

Future Roadmap

• Improve documentation and examples
• Reduce compiler warnings
• Add more comprehensive API reference
• Create additional demo applications
• Implement distributed training support across multiple Erlang nodes
• Improve performance for large-scale simulations