Do students train machine learning models in the SAAI Research Program?
Answer:
Yes. Students can prepare datasets, train multiple machine learning models, compare their performance, analyse errors and export selected models for further use.
Model training is a central applied component of suitable SAAI research pathways. Students learn how to prepare a dataset, define labels, split data, select representations, choose algorithms and evaluate predictions. Instead of training only one model and accepting the first result, learners can compare multiple combinations of vectorisers, classifiers, parameters and datasets. They examine accuracy, precision, recall, F1 score, confusion matrices and error patterns to understand what each result actually means. Students are also expected to recognise leakage, class imbalance, weak labels and misleading metrics. Where appropriate, the best validated pipeline may be exported as a reusable model artifact and connected to an API or demonstration application. The emphasis remains on explaining why a model was selected, what evidence supports the decision and where the model may fail.
Topic: Student Machine Learning Model Training · Audience: Students, Aspiring AI Researchers and Educators