Table 1 Comparison between the levels of accuracy, complexity, and interpretability offered by the machine learning algorithms used herein, namely, polynomial regression (PR), LASSO, random forest (RF), artificial neural network (ANN).

From: Predicting the Young’s Modulus of Silicate Glasses using High-Throughput Molecular Dynamics Simulations and Machine Learning

ML algorithms

Coefficient of determination R2

Complexity

Interpretability

Training set

Test set

PR

0.975

0.970

Low (9)

High

LASSO

0.971

0.966

Low (8)

High

RF

0.991

0.965

High (200)

Intermediate

ANN

0.980

0.975

Intermediate (20)

Low

  1. The level of accuracy is described by the coefficient of determination (R2) for the training and test sets. The complexity is described in parenthesis by the number of non-zero parameters in PR and LASSO, the number of trees in RF, and the product of the number of inputs, neurons, and adjustable parameters per neuron in ANN. The “interpretability” describes the degree to which a human can understand the outcome produced by each model.
Back to article page