Machine learning’s ability to perform intellectually demanding tasks across various fields, materials science included, has caused it to receive considerable attention. Many believe that it could be used to unlock major time and cost savings in the development of new materials.

The growing demand for the use of machine learning to derive fast-to-evaluate surrogate models of material properties has prompted scientists at the National Institute for Materials Science in Tsukuba, Japan, to demonstrate that it could be the key driver of the “next frontier” of materials science in recently published research.  

Insufficient Materials Data

To learn, machines rely on processing data using both supervised and unsupervised learning. 

With no data, however, there is nothing to learn from. 

Unfortunately, potential technological advances in machine learning and its potential applications in materials science are not being fully exploited due to a considerable lack of volume and diversity of materials data. This, the Japanese researchers believe, is greatly stifling progress.

However, a machine learning framework known as “transfer learning” is said to have great potential in being able to overcome the problem of a relatively small data supply. This framework relies on the concept that various property types – for example, physical, electrical, chemical, and mechanical – are physically interrelated. 

To successfully predict a target property from a limited supply of training data, the researchers used models of related proxy properties that have been pretrained using sufficient data. These models are then able to capture common features that are relevant to the target task.

This repurposing of machine-acquired features on the target task has demonstrated high prediction performance even when the data sets are very small

Facilitating the Widespread Use of Transfer Learning

To facilitate the widespread use and boost the power of transfer learning, the Japanese researchers created their own pre-trained model library, XenonPy.MDL. 

In its first release, the library is comprised of more than 140,000 pre-trained models for various properties of small molecules, polymers, and inorganic crystalline materials. Along with these models, the researchers provide literature that describes some of their most outstanding successes of applying transfer learning in varying scenarios. Examples such as being able to build models using the framework in conjunction with only dozens of pieces of materials data nicely hammer home the effectiveness of the transfer learning framework.

The researchers also highlight how they discovered that transfer learning transcends across different disciplines of materials science, such as underlying bridges between organic and inorganic chemistry. 

Indispensable to Machine Learning-Centric Workflows

Although transfer learning is being used often across various fields of machine learning, its use in materials science is still lacking. 

Furthermore, the limited availability of openly accessible big data will likely continue in the near future, the researchers say, due to a distinct lack of incentives for data sharing, a problem partly caused by the conflicting goals of stakeholders across academia, industry, and public and government organizations. This means that the relevance of transfer learning will only grow and contribute further to the success of machine learning-centric workflows in materials science.

The post Using Transfer Learning to Overcome the Barriers Facing Machine Learning in Materials Science appeared first on Artificial Intelligence.

Medical imaging is among the most popular application of AI and machine learning, and with good reason. Computer vision algorithms are naturally adept at spotting anomalies experts sometimes miss, in the process reducing wait times and lightening clinical workloads. Perhaps that’s why although the percentage of health care organizations that have adopted AI remains relatively low (22%) globally, the majority of practitioners (77%) believe the technology is important to the medical imaging field as a whole.

Unsurprisingly, data scientists have devoted outsize time and attention to developing AI imaging models for use in health care systems, a few of which Google scientists detail in a paper accepted to this week’s NeurIPS conference in Vancouver. In “Transfusion: Understanding Transfer Learning for Medical Imaging,” coauthors hailing from Google Research (the R&D-focused arm of Google’s business) investigate the role transfer learning plays in developing image classification algorithms.

In transfer learning, a machine learning algorithm is trained in two stages. First, there’s retraining, where the algorithm is generally trained on a benchmark data set representing a diversity of categories. Next comes fine-tuning, where it is further trained on the specific target task of interest. The pretraining step helps the model to learn general features that can be reused on the target task, boosting its accuracy.

According to the team, transfer learning isn’t quite the end-all, be-all of AI training techniques. In a performance evaluation that compared a range of model architectures trained to diagnose diabetic retinopathy and five different diseases from chest x-rays, a portion of which were pretrained on an open source image data set (ImageNet), they report that transfer learning didn’t “significantly” affect performance on medical imaging tasks. Moreover, a family of simple, lightweight models performed at a level comparable to the standard architectures.

In a second test, the team studied the degree to which transfer learning affected the kinds of features and representations learned by the AI models. They analyzed and compared the hidden representations (i.e., representations of data learned in the model’s latent portions) in the different models trained to solve medical imaging tasks, computing similarity scores for some of the representations between models trained from scratch and those pretrained on ImageNet. The team concludes that for large models, representations learned from scratch tended to be much more similar to each other than those learned from transfer learning, while there was greater overlap between representation similarity scores in the case of smaller models.

To rectify these and other issues, the team proposes a hybrid approach to transfer learning, where instead of reusing the full model architecture, only a portion of is resused and the rest is redesigned to better suit the target task. They say that it confers most of the benefits of transfer learning while further enabling flexible model design. “Transfer learning is a central technique for many domain,” wrote Google Research scientists Maithra Raghu and Chiyuan Zhang in a blog post. “Many interesting open questions remain, [and we] look forward to tackling these questions in future work.”

The work comes shortly after Google detailed an AI capable of classifying chest X-rays with human-level accuracy. In another recent study, teams from the tech giant claimed to have developed a machine learning model that detects 26 skin conditions as accurately as dermatologists and a lung cancer detection AI that outperformed six human radiologists.

The post Google proposes hybrid approach to AI transfer learning for medical imaging appeared first on Artificial Intelligence.