🍃Prebuilt Pipeline
Let's look at prebuilt pipelines first!
TPOT is an AutoML pipeline that uses genetic algorithm to select the best pipeline for the data.
AutoKeras builds the optimal neural network model for the data, it does parameter tuning and model selection using Keras Tuner.
MLJar selects the best model from classical machine learning models and neural network models. Besides, it also generates detailed reports and visualization for each model.
The experiments utilized the same datasets as those in the hyperparameter optimization (HPO) experiments. The evaluation results, summarized in Table 4.1, compare the performance of TPOT, AutoKeras, and MLJar, as well as their results against the best HPO performance.
When using these prebuilt pipelines, the execution time is expected to be longer than that of HPO, as more models and parameters need to be selected.
Last updated