Neural pattern training with SONA (Self-Optimizing Neural Architecture), MoE (Mixture of Experts), and EWC++ for knowledge consolidation. Use when: pattern learning, model optimization, knowledge transfer, adaptive routing. Skip when: simple tasks, no learning required, one-off operations.
Train and optimize neural patterns using SONA, MoE, and EWC++ systems.
| Component | Purpose | Performance |
|---|
| SONA | Self-optimizing adaptation | <0.05ms |
| MoE | Expert routing | 8 experts |
| HNSW | Pattern search | 150x-12,500x |
| EWC++ | Prevent forgetting | Continuous |
| Flash Attention | Speed | 2.49x-7.47x |
npx claude-flow neural train --model-type moe --epochs 10
npx claude-flow neural status
npx claude-flow neural patterns --type all
npx claude-flow neural predict --input "task description"
npx claude-flow neural optimize --target latency