neural-training
// 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.
$ git log --oneline --stat
stars:18,689
forks:3.6k
updated:February 28, 2026
SKILL.mdreadonly
SKILL.md Frontmatter
nameneural-training
descriptionNeural 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.
Neural Training Skill
Purpose
Train and optimize neural patterns using SONA, MoE, and EWC++ systems.
When to Trigger
- Training new patterns
- Optimizing agent routing
- Knowledge consolidation
- Pattern recognition tasks
Intelligence Pipeline
- RETRIEVE — Fetch relevant patterns via HNSW (150x-12,500x faster)
- JUDGE — Evaluate with verdicts (success$failure)
- DISTILL — Extract key learnings via LoRA
- CONSOLIDATE — Prevent catastrophic forgetting via EWC++
Components
| 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 |
Commands
Train Patterns
npx claude-flow neural train --model-type moe --epochs 10
Check Status
npx claude-flow neural status
View Patterns
npx claude-flow neural patterns --type all
Predict
npx claude-flow neural predict --input "task description"
Optimize
npx claude-flow neural optimize --target latency
Best Practices
- Use pretrain hook for batch learning
- Store successful patterns after completion
- Consolidate regularly to prevent forgetting
- Route based on task complexity