Why big data and compute are not necessarily the path to big materials science

Abstract

Applied machine learning has rapidly spread throughout the physical sciences. In fact, machine learning-based data analysis and experimental decision-making have become commonplace. Here, we reflect on the ongoing shift in the conversation from proving that machine learning can be used, to how to effectively implement it for advancing materials science. In particular, we advocate a shift from a big data and large-scale computations mentality to a model-oriented approach that prioritizes the use of machine learning to support the ecosystem of computational models and experimental measurements. We also recommend an open conversation about dataset bias to stabilize productive research through careful model interrogation and deliberate exploitation of known biases. Further, we encourage the community to develop machine learning methods that connect experiments with theoretical models to increase scientific understanding rather than incrementally optimizing materials. Moreover, we envision a future of radical materials innovations enabled by computational creativity tools combined with online visualization and analysis tools that support active outside-the-box thinking within the scientific knowledge feedback loop.

Probe microscopy is all you need *

“…A number of the leading scientists now argue that artificial intelligence (AI) needs to move away from ever-increasing dataset and network sizes, to essentially 'do more with less [58] by using small data sets and physical priors. Similar arguments have been made in the physical science field [59,60], typically in cognizance of rich systems of prior knowledge and inferential biases and extremely small experimental budgets. This convergence between ML and physical fields is, in our opinion, a significant aspect of the last three years.…”

Machine learning using multi-modal data predicts the production of selective laser sintered 3D printed drug products

“…A concerted effort has been undertaken by scientists across different disciplines to address the need for ML for small datasets (Elbadawi et al, 2021a). There is undoubtedly a fallacy surrounding the idea that large data is needed for feasible predictions (Fujinuma et al, 2022). An effective model can indeed be built with only a few data instances (Baskin, 2019).…”

In Pursuit of the Exceptional: Research Directions for Machine Learning in Chemical and Materials Science

“…Limitations of data-driven strategies have been noted in the literature, with the need for more data or higher-quality data being stressed . Algorithmic performance can also depend on the initial data set (the “cold start” problem), and available data sets often exhibit sampling biases .…”

“…Classically, this was done by devising physical models in terms of the relevant variables and the admissible functional forms of their interactions. Physics-based computer simulations serve a similar role, although the examples above indicate their limits for exceptional materials. , We focus purely on data-driven approaches. Strategies of physics-informed machine learning − are one approach for this problem.…”