Authors: Samuel Yuan, S.V. Dordevic

Published on: January 31, 2024

Impact Score: 8.45

Arxiv code: Arxiv:2402.00198

Summary

• What is new: Implementation of conditioning in generative models for the first time to discover new families of High-Temperature Superconductors (HTSs).
• Why this is important: Existing machine learning models could predict new superconductors within known families but couldn’t generate entirely new families of superconductors.
• What the research proposes: Developed SuperDiff, a Denoising Diffusion Probabilistic Model with Iterative Latent Variable Refinement conditioning, to control the generation process and discover new HTS families.
• Results: Successfully generated completely new families of hypothetical superconductors, potentially accelerating the discovery of new superconductors.

Technical Details

Technological frameworks used: Denoising Diffusion Probabilistic Models (DDPM)

Models used: SuperDiff with Iterative Latent Variable Refinement (ILVR) conditioning

Data used: nan

Potential Impact

Materials science companies, semiconductor industry, energy sector

Want to implement this idea in a business?

We have generated a startup concept here: SuperAISearch.