Machine learning that evolves on the mesh. Neural networks that grow, adapt, and learn in real-time across distributed BEAM nodes.
Libraries for building and evolving neural networks on the BEAM.
Topology and Weight Evolving Artificial Neural Networks. Networks that grow and adapt with Liquid Time-Constant neurons.
TWEANN Neuroevolution, enhanced with a conglomerate of evolving LTC Neural Networks, enables emerging Evolutionary Strategies through meta-learning.
The vision: distributed neural evolution across the Macula mesh. Populations that evolve across nodes, with fitness computed in parallel.
The network topology evolves automatically. No need to manually design layer structures.
Networks continue learning and adapting as environments change.
Evolution finds the smallest network that solves the problem. No over-parameterization.
Evolutionary algorithms discover creative solutions that gradient descent might miss.
Game AI, robotics, and decision-making systems that adapt to their environment.
LTC networks excel at temporal patterns in sensor data, finance, and monitoring.
Compact networks that run efficiently on resource-constrained devices.
Continuous control for robotics, drones, and process automation.
Neural network computations run as Rust NIFs for maximum performance, while evolution and coordination happen in Erlang/Elixir for fault tolerance.
Add Macula ML to your project and let evolution design your neural networks.
Read the Docs