Regression, classification, representation learning and experiment selection, with materials examples for training, validation and uncertainty analysis.
No matching tutorials.
Compare linear regression, regularization, kernel methods, tree ensembles, and boosting through their assumptions and fitting behavior. Use residuals, cross-validation, and suitable error metrics to select a model for numerical prediction.
Follow a neural network from weighted inputs and activations to loss functions, backpropagation, and optimization. Connect these operations to convolutional, recurrent, and attention models, with practical guidance on training and small datasets.
Construct useful material descriptors, encode categorical processing conditions, and handle interactions and missing values. Learn how feature selection and training-only preprocessing keep a small-data model interpretable and its evaluation consistent.
Compare logistic regression, support vector machines, trees, ensembles, nearest neighbors, and naive Bayes. Read class probabilities and confusion matrices, then choose metrics and validation procedures that match the cost of classification errors.
Build a small-data regression workflow for fiber mechanics, from specimen features and preprocessing to model comparison. Examine cross-validation, bootstrap variation, and feature importance to connect numerical predictions with material descriptors.