Talos provides access to sets of templates consisting of the assets required by Scan() i.e. datasets, parameter dictionaries, and input models. In addition, talos.templates.pipelines consist of ready pipelines that combined the assets into a Scan() experiment and run it. These are mainly provided for educational, testing, and development purposes.
Each category of templates consists at least assets based on four popular machine learning datasets:
- Wisconsin Breast Cancer | dataset info
- Cervical Cancer Screening | dataset info
- Iris | dataset info
- Titanic Survival | dataset info
In addition, some categories (e.g. datasets) include additional templates. These are listed below and can be accessed through the corresponding namespace without previous knowledge.
Datasets are preprocessed so that they can be used directly as inputs for deep learning models. Datasets are accessed through talos.templates.datasets. For example:
talos.templates.datasets.breast_cancer()- breast_cancer
- cervical_cancer
- icu_mortality
- telco_churn
- titanic
- iris
- mnist
Params consist of an indicative and somewhat meaningful parameter space boundaries that can be used as the parameter dictionary for Scan() experiments. Parameter dictionaries are accessed through talos.templates.params. For example:
talos.templates.params.breast_cancer()- breast_cancer
- cervical_cancer
- titanic
- iris
Models consist of Keras models that can be used as an input model for Scan() experiments. Models are accessed through talos.templates.models. For example:
talos.templates.models.breast_cancer()- breast_cancer
- cervical_cancer
- titanic
- iris
Pipelines are self-contained Scan() experiments where you simply execute the command and an experiment is performed. Pipelines are accessed through talos.templates.pipelines. For example:
scan_object = talos.templates.pipelines.breast_cancer()- breast_cancer
- cervical_cancer
- titanic
- iris