Over a million developers have joined DZone.
{{announcement.body}}
{{announcement.title}}

Using AI to Help Create New Materials

DZone's Guide to

Using AI to Help Create New Materials

A team led by researchers at MIT have developed a new AI-driven approach that would do a lot of the research legwork and propose new "recipes" for new materials.

· AI Zone ·
Free Resource

Insight for I&O leaders on deploying AIOps platforms to enhance performance monitoring today. Read the Guide.

The Materials Genome Initiative is something I've touched on a few times over the years, and it and similar projects have helped in not only better understanding the health and lifecycle of materials but also in developing new computational tools for the design of new materials. While these tools are a great help, much of the development has still depended upon manual, labor-intensive processes.

A team led by researchers at MIT have developed a new AI-driven approach that would do a lot of the research legwork and propose new "recipes" for new materials.

"Computational materials scientists have made a lot of progress in the 'what' to make-what material to design based on desired properties," the authors say. "But because of that success, the bottleneck has shifted to, 'Okay, now how do I make it?'"

Computational Research

The team has "fed" the system on millions of research papers that have allowed scientists and engineers to enter in the name of a target material and various other criteria they want from it, and be returned a number of suggested recipes.

The algorithm is capable of analyzing each paper and deducing the parts that contain useful information for the creation of materials. What's more, it can then classify the words according to their role within each recipe.

The system was trained using a mixture of supervised and unsupervised machine learning, due in large part to the lack of annotated datasets available to the team. This forced them to do the annotation themselves of a small sample. They then built on this via the Google Word2vec algorithm, which is able of understanding the context within which words occur.

Through this, they were able to produce a suitably significant training set to drill the algorithm on, before then putting it through its paces. During the testing phase, the system was able to accurately identify paragraphs containing a recipe 99% of the time.

The next stage will be to develop the algorithm further to both improve its accuracy but also to enable it to make further generalizations about the structure of material recipes. This will be a crucial step in the process of eventually using AI to propose new recipes itself.

TrueSight is an AIOps platform, powered by machine learning and analytics, that elevates IT operations to address multi-cloud complexity and the speed of digital transformation.

Topics:
materials science ,ai ,automation ,algorithm ,machine learning

Published at DZone with permission of

Opinions expressed by DZone contributors are their own.

{{ parent.title || parent.header.title}}

{{ parent.tldr }}

{{ parent.urlSource.name }}