| About this Abstract |
| Meeting |
2023 TMS Annual Meeting & Exhibition
|
| Symposium
|
Computational Discovery and Design of Materials
|
| Presentation Title |
Crystal to PNG (xtal2png): A Screening Tool to Accelerate Domain Transfer from State-of-the-art Image-processing Models to Materials Informatics and a Case Study on Denoising Diffusion Probabilistic Models |
| Author(s) |
Sterling G. Baird, Kevin M. Jablonka, Michael D. Alverson, Hasan M. Sayeed, Faris Khan, Colton Seegmiller, Berend Smit, Taylor D. Sparks |
| On-Site Speaker (Planned) |
Sterling G. Baird |
| Abstract Scope |
The latest advances in machine learning are often in natural language such as with long short-term memory networks (LSTMs) and transformers or image processing such as with generative adversarial networks (GANs), variational autoencoders (VAEs), and guided diffusion models. Using xtal2png to encode/decode crystal structures via grayscale PNG images is akin to making/reading a QR code for crystal structures. This allows you, as a materials informatics practitioner, to get streamlined results for new state-of-the-art image-based machine learning models applied to crystal structure. To showcase xtal2png, we apply it to several denoising diffusion probabilistic models (DDPMs), a recent state-of-the-art model in the image-processing domain. We also provide results using our recently developed generative benchmarking framework (https://matbench-generative.readthedocs.io/) and compare with CDVAE, a state-of-the-art generative model for crystal structures. The xtal2png Python package is open-source and documentation can be found at https://xtal2png.readthedocs.io/ |
| Proceedings Inclusion? |
Planned: |
| Keywords |
Machine Learning, Computational Materials Science & Engineering, Other |