Examples - Basic
Workflows don't have to be complex to get the job done. But even flows with only 10 actors already help documenting your work, preserving all steps in volved in generating the results. These flows are a good introduction into the many actors that ADAMS already has on offer.
Note: the videos should only be considered educational, as some of the concepts in ADAMS have changed over time. E.g., global actors are now called callable, since they can appear in different scopes within the flow. Also, SingleFileSupplier and MultiFileSupplier got merged into the FileSupplier actor.
Hello world flow
Absolute simple flow that merely displays the Hello World string.
Cross-validating a single classifier on one dataset
Cross-validates a classifier (J48) on a single UCI dataset.
Dynamic evaluation of multiple classifiers on multiple datasets
Reads in set ups of WEKA classifiers stored in a text file and evaluates each against all datasets in a directory.
Read in text file with classifier set ups, line by line, and store set up in in variable attached to the global actor (see 5)
Iterate over all datasets in directory and load them, one after the other
Drop datasets that don't meet required capabilities
Cross-validate classifier defined in 5 on datasets and output results
The classifier definition used in cross-validation, with variable attached to set up (gets updated when the variable changes its value)
Experiment on arbitrary datasets
Adds datasets located in a directory to a pre-defined WEKA experiment, runs the experiment and displays the evaluation.
Lists datasets in specified directory
Executes the pre-defined WEKA experiment (containing only classifier set ups, no datasets) using the datasets arriving at its input
Performs the evaluation of the experiment
Classifier ranking of parameter sweep
Generates a parameter sweep for multiple parameters of WEKA classifiers, ranks these using 2-fold cross-validation and subjects the top 3 to a proper evaluation.
Generates the parameter sweep of 3 parameters: ridge of LinearRegression, # of components of PLS filter and type of PLS algorithm
Outputs all the generated setups
Ranks the performance of the classifiers using 2-fold cross-validation, outputs the top 3
Displays the set ups of the top 3 classifiers
The global training data