skills.rest
learningmatter-mit
learningmatter-mitOFFICIAL
Explore all skills by this author →
WebsiteGitHub

ml-committee-uncertainty

Estimate epistemic uncertainty of MACE MLIPs from committee disagreement.

144|21|Updated Jan 8, 2026
Search Tags:#active learning#uncertainty quantification#mace#ml interatomic potentials#DFT labeling#ensemble committee#atomistic MD
One-click install
npx skills add https://github.com/learningmatter-mit/AtomisticSkills --skill ml-committee-uncertainty
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: ml-committee-uncertainty
Source: https://github.com/learningmatter-mit/AtomisticSkills/tree/main/.agents/skills/ml-committee-uncertainty
Command: npx skills add https://github.com/learningmatter-mit/AtomisticSkills --skill ml-committee-uncertainty

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires numpy, matplotlib, ase, mace-torch, and includes scripts (resource) components.

What problem does it solve?

This Skill quantifies epistemic uncertainty in a MACE machine-learning interatomic potential by measuring disagreement across a committee of independently trained models, so you can identify out-of-distribution structures that should be verified or labeled with DFT.

Core Features & Use Cases

  • Committee-based uncertainty (energy and forces): Computes energy standard deviation (meV/atom) and force variance/std (meV/Å) from multiple MACE checkpoints.
  • Automatic DFT flagging: Flags structures whose uncertainty exceeds configurable energy/force thresholds and saves them for labelling.
  • Active-learning ready outputs: Writes per-structure summaries (uncertainty_summary.json), flagged structure files, and uncertainty distributions for analysis.
  • Use Case: Run uncertainty screening on an MD trajectory to find when your system drifts beyond the MLIP training distribution, then send the flagged frames to a DFT labelling workflow.

Quick Start

Run committee inference on your trajectory by providing the input structures file and at least three MACE checkpoint paths to produce flagged structures for DFT verification.

Frequently Asked Questions about ml-committee-uncertainty

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I flag uncertain MACE structures for DFT labeling?▼

You can flag uncertain MACE structures by computing energy and force disagreement across a committee of at least three independently trained checkpoints, then applying thresholds to identify out-of-distribution configurations for DFT verification.

What is committee-based uncertainty quantification for machine-learning interatomic potentials?▼

Committee-based uncertainty quantification estimates epistemic uncertainty by measuring the standard deviation of energies and forces across multiple independently trained MLIP models to detect out-of-distribution structures.

How do I detect out-of-distribution frames in an atomistic MD trajectory?▼

You can detect out-of-distribution frames in an atomistic MD trajectory by running committee inference with multiple MACE models to measure energy and force variance, then exporting structures that exceed your uncertainty thresholds.

Do I need multiple MACE checkpoints to calculate MLIP uncertainty?▼

Yes, you need at least three independently trained MACE checkpoints to calculate MLIP uncertainty, because the method relies on measuring energy and force disagreement across the committee of models.

What outputs do I get from MLIP active learning uncertainty screening?▼

You get an uncertainty_summary.json file containing per-structure statistics, flagged structure files for DFT labeling, and uncertainty distributions for analyzing your active learning workflow.

Can I use this committee uncertainty method with ASE and MACE-torch?▼

Yes, this method works with ASE and MACE-torch, requiring numpy and matplotlib alongside these dependencies to process atomistic structures and visualize uncertainty distributions for your active learning workflow.

Related Skills

Explore More Education & Research Skills→
learningmatter-mitlearningmatter-mit
144

general-property-units

Define ASE-standard energy, force, and stress units across MLIP and DFT toolchains.

Official
Basic
learningmatter-mitlearningmatter-mit
144

mat-sample-pes-by-md

Sample off-equilibrium atomic configurations via MD and latent-feature clustering.

Official
Advanced
jkitchinjkitchin
34

python-ase

Build ASE atomistic structures and run simulations with calculators.

Community
Advanced
baratadiegobaratadiego
2

model-validation-and-uncertainty

Compute performance metrics and diagnostic plots for ecological predictive models.

Community
Advanced