Computational Prediction of Critical Temperatures of Superconductors Based on Convolutional Gradient Boosting Decision Trees

Abstract

Superconductors have been one of the most intriguing materials since they were discovered more than a century ago. However, superconductors at room temperature have yet to be discovered. On the other hand, machine learning and especially deep learning has been increasingly used in material properties prediction and discovery in recent years. In this paper, we propose to combine the deep convolutional neural network (CNN) model with fully convolutional layers for feature extraction with gradient boosting decision tree (GBDT) for superconductors critical temperature (T c) prediction. Our prediction model only uses the elemental property statistics of the materials as original input and learns a hierarchical representation of superconductors using convolutional layers. Computational experiments showed that our convolutional gradient boosting decision tree (ConvGBDT) model achieved the state-of-the-art results on three superconductor data sets: DataS, DataH, and DataK. By visually comparing the raw elemental feature distribution and the learned feature distribution, it is found that the convolutional layers of our ConvGBDT can learn features that can more effectively distinguish cuprate and iron-based superconductors. On the other hand, the GBDT part of our ConvGBDT model can learn the sophisticated mapping relationship between extracted features and the critical temperatures to obtain good prediction performance.

Upper limit of the transition temperature of superconducting materials

“… 25 , 26 , 27 ML models using different algorithms were trained to predict the existence of superconductivity and the T c of superconductors. 28 , 29 , 30 , 31 , 32 , 33 , 34 , 35 , 36 , 37 , 38 , 39 , 40 , 41 , 42 , 43 , 44 , 45 , 46 , 47 , 48 , 49 , 50 , 51 , 52 , 53 , 54 Progress has been made in several areas, such as how T c varies with doping, 28 , 29 , 30 , 31 , 32 , 33 , 34 , 35 the descriptors indicating superconducting mechanism, 36 , 37 , 38 , 39 structural factors affecting T c , 43 , 44 and candidates of new high- T c superconductors. 46 , 51 So far, ML models predicting T c have yielded good predictive scores.…”

The CTCN-LightGBM Joint Model for Industrial Balanced Loading Prediction

“…Lu et al [26] integrate the long short-term memory with the gradient boosting machine to predict end-to-end inferences. Dan et al [27] combine a convolutional neural network with the gradient-boosting decision tree for temperature prediction. Also, Ju et al [28] propose a convolutional neural network and light-GBM model to predict wind power.…”

Machine Learning Prediction of Superconducting Critical Temperature through the Structural Descriptor

“…Recently, first-principles calculations, molecular dynamics, and other multiscale computational methods have been applied for designing a wide range of materials with attractive superconducting properties. − These methods are usually used to examine the mechanism of superconductivity, while the T c values of superconductors are generally measured by experiments instead of theoretical predictions because of the time-consuming procedures and undesirable accuracies . The data-driven scientific developments in artificial intelligence (AI) algorithms, particularly the machine learning (ML) and deep learning techniques, − may provide alternative approaches to predict the T c values of potential superconducting materials. − …”

“…Stanev et al recently used the Magpie descriptor to characterize 12,000 known superconductors using a 132-dimensional vector and reached a regression accuracy of 88%. Among these data-driven efforts in search for new superconductors, the role of atomic structures of materials has not yet been explicitly constructed in previous models, − which may be crucial to the prediction accuracy and reliability, as well as the ultimate structure–property relationship goal of the condensed matter theory. Generally, the descriptors derived from elemental property statistics, such as the standard deviation and the mean of different elemental properties, were utilized to predict T c values in recent ML-based prediction methods. − Nevertheless, when atomic structure information of superconductors was considered during the prediction processes, the ML method would have to learn rotational and translational invariance for every data set, which would presumingly increase the model accuracy.…”