---

Machine learning gets better at admitting uncertainty on messy, interconnected data

---

Imagine a loan officer who checks not just your credit score, but also how your score relates to your neighbors' — not to judge you, but to know how confident she should be in her decision. GRAPHLCP does that for neural networks operating on graph-shaped data.

This means predictions on networks (social graphs, knowledge bases, recommendation systems) can now come with honest uncertainty bounds, letting you know when the model is guessing versus genuinely confident.

---