Rational design of high-entropy ceramics based on machine learning – A critical review

Accelerating the discovery of high‐entropy hexaborides by data‐driven prediction: From equimolar to non‐equimolar

“…To study the synthesizability of HEB 6 , seven commonly used ML models, including k-nearest neighbors (KNN), decision trees (DT), logistic regression (LR), Gaussian Naive Bayes (GNB), support vector machine with a lin-ear kernel (SVM.linear), support vector machine with a polynomial kernel (SVM.poly), and support vector machine with a radial basis function kernel (SVM.rbf), were applied. 31 The training dataset was collected from the results of high-throughput synthesis experiments on 100 equimolar quinary HEB 6 samples (see Table S1) and 20 features based on the fundamental parameters of constituent AEEs, REEs, and hexaborides (see Table 1). It should be noted that the single-and multi-phase HEB 6 samples were signed as "1" and "0", respectively, in the ML.…”

Advancements in Microwave Absorption Motivated by Interdisciplinary Research

“…Figure a illustrates the ML workflow for forward prediction, which comprises data collection, database organization, feature engineering, model training, application, and experimental validation. [ 158 ] Data collection is arguably the most important as the performance of the ML model depends heavily on the quality and quantity of the data. Data can be obtained from publicly available resources or be self‐generated, e.g., by carrying out experiments or numerical simulations.…”

Machine learning‐based crystal structure prediction for high‐entropy oxide ceramics

“…According to the obtaining pathways, the input features can be classified into three categories, including the atomic/precursory-based descriptors, the experienceindependent features, and the characteristics acquired by high-throughput simulation. 41 Here, in order to avoid time-consuming and expensive high-throughput calculations, the first two types of features were selected to form the feature space where the first type descriptors consist of anion-to-cation radius ratio (𝑟 𝐴 ∕𝑟 𝐶 ), difference in Pauling electronegativity (Δ𝑥 Pauling ), 42 difference in Mulliken electronegativity (Δ𝑥 Mulliken ), 43 entropy of mixing (Δ𝑆 mix ) 44 as well as atomic size mismatch (Δ𝛿), 42 and the corresponding calculation formulas are as follows 31,45 :…”

“…In this work, the dataset consists of 112 experimental data, which were summarized by Akrami et al., 2 as can be seen in Table S1. According to the obtaining pathways, the input features can be classified into three categories, including the atomic/precursory‐based descriptors, the experience‐independent features, and the characteristics acquired by high‐throughput simulation 41 . Here, in order to avoid time‐consuming and expensive high‐throughput calculations, the first two types of features were selected to form the feature space where the first type descriptors consist of anion‐to‐cation radius ratio ($${r_A}/{r_C}$$), difference in Pauling electronegativity ($$\Delta {x_{{\mathrm{Pauling}}}}$$), 42 difference in Mulliken electronegativity ($$\Delta {x_{{\mathrm{Mulliken}}}}$$), 43 entropy of mixing ($$\Delta {S_{{\mathrm{mix}}}}$$) 44 as well as atomic size mismatch ($$\Delta \delta $$), 42 and the corresponding calculation formulas are as follows 31,45 : $$$$\begin{equation}{r_A}/{r_C} = \frac{{{\mathrm{Radius\;of\;anion}}}}{{{\mathrm{Radius\;of\;cation}}}},\end{equation}$$$$ $$$$\begin{equation}\Delta {x_{{\mathrm{Pauling}}}} = \sqrt {\mathop \sum \limits_{i = 1}^n {c_i}{{( {{x_i} - {{\bar x}_p}} )}^2}} ,\end{equation}$$$$…”