SynCoTrain: A Dual Classifier PU-learning Framework for Synthesizability Prediction

TL;DR

This work establishes SynCoTrain as a reliable tool for predicting synthesizability while balancing dataset variability and computational efficiency, and highlights the potential of co-training to advance high-throughput materials discovery and generative research.

Abstract

Material discovery is a cornerstone of modern science, driving advancements in diverse disciplines from biomedical technology to climate solutions. Predicting synthesizability, a critical factor in realizing novel materials, remains a complex challenge due to the limitations of traditional heuristics and thermodynamic proxies. While stability metrics such as formation energy offer partial insights, they fail to account for kinetic factors and technological constraints that influence synthesis outcomes. These challenges are further compounded by the scarcity of negative data, as failed synthesis attempts are often unpublished or context-specific. We present SynCoTrain, a semi-supervised machine learning model designed to predict the synthesizability of materials. SynCoTrain employs a co-training framework leveraging two complementary graph convolutional neural networks: SchNet and ALIGNN. By iteratively exchanging predictions between classifiers, SynCoTrain mitigates model bias and enhances generalizability. Our approach uses Positive and Unlabeled (PU) Learning to address the absence of explicit negative data, iteratively refining predictions through collaborative learning. The model demonstrates robust performance, achieving high recall on internal and leave-out test sets. By focusing on oxide crystals, a well-characterized material family with extensive experimental data, we establish SynCoTrain as a reliable tool for predicting synthesizability while balancing dataset variability and computational efficiency. This work highlights the potential of co-training to advance high-throughput materials discovery and generative research, offering a scalable solution to the challenge of synthesizability prediction.

Figures (10)

• Figure 1: Overview of the Workflow in SynCoTrain a) The PU data is passed to two distinct PU classifiers, each learning from a different view of the data. Each classifier labels unlabeled data points as positive or negative. The new labels from each PU classifier are used to expand the positive class for retraining the other classifier. b) After co-training steps, each unlabeled data point receives a prediction score from each PU classifier. An average of these scores is calculated for each data point, and a cutoff is applied to produce a label. All the data, now labeled, are used to train a final classifier to predict synthesizability. c) The PU learning process. Positive, negative, and unlabeled data are depicted as green, red, and gray circles respectively. Each run starts with training a classifier, with a randomly chosen subset of the unlabeled data used as the negative class. Labels are predicted for the remaining unlabeled data, and final scores are computed by averaging these predictions.
• Figure 2: Recall progression per iteration for both co-training series
• Figure 3: Synthesizability score distribution for Iteration '0' for the a) ALIGNN0 series and b) SchNet0 series.
• Figure 4: Synthesizability score distribution for Iteration '2' for the a) ALIGNN0 series and b) SchNet0 series.
• Figure 5: Label distribution after averaging scores