Expand description
Neural Network Primitives for AetherShell
Provides lightweight neural network building blocks for in-shell learning, particularly useful for evolving multi-agent communication protocols and consensus mechanisms.
Structs§
- Attention
Layer - Attention mechanism for agent communication
- Consensus
Network - Consensus network for distributed decision making
- Dense
Layer - A dense (fully connected) layer
- Message
Codec - Message encoder/decoder for agent communication
- Network
Registry - Registry for managing neural networks in the shell
- Neural
Network - A complete neural network
- RNNCell
- A recurrent neural network cell (simple RNN)
Enums§
- Activation
- Activation functions for neural network layers