Icon

05_​Weak_​Supervision_​on_​the_​Adult_​dataset

Weak Supervision on the Adult dataset

This workflow shows how to use the Weak Label Model Learner and Predictor nodes to aggregate sources of weak supervision such as weak models or simple rules into a single strong source that can be used to train various models.
The models trained here are Gradient Boosted Trees, Logistic Regression and Deep Learning.
In the final step of the workflow these models are applied to unseen data and the results are visualized with the Binary Classification Inspector that allows to compare the performance of different models in an interactive view.

The data used is a preprocessed version of the Adult dataset (https://archive.ics.uci.edu/ml/datasets/Adult)

Collect weak supervision sources Combine multiple weak sources into onestrong source Learn from the strong source Apply the learned models to new data Visualize model performance Read data Learn tocombine weaksourcesCombine weaksources intoa single strongsupervisionsource Create probabilitydistributions frompredicted probabilitiesWeak modelSplit probabilitiesof strong source intoindividual columnsfor deep learningRead unlabeledtraining dataRead newunlabeled dataExtract classesfrom probabilitiesNormalizenumericalfeaturesApply same normalizationas during learningWeak LabelModel Learner Weak Label ModelPredictor Nominal ProbabilityDistribution Creator Joiner Weak Source PMML Reader Decision TreePredictor Joiner Gradient BoostedTrees Learner Gradient BoostedTrees Predictor LogisticRegression Learner Logistic RegressionPredictor Keras NetworkLearner Nominal ProbabilityDistribution Splitter Table Reader Table Reader Keras NetworkExecutor Weak Source Define Network Column Rename(Regex) Normalizer Normalizer (Apply) Binary ClassificationInspector Decision TreePredictor Collect Predictions Collect weak supervision sources Combine multiple weak sources into onestrong source Learn from the strong source Apply the learned models to new data Visualize model performance Read data Learn tocombine weaksourcesCombine weaksources intoa single strongsupervisionsourceCreate probabilitydistributions frompredicted probabilitiesWeak modelSplit probabilitiesof strong source intoindividual columnsfor deep learningRead unlabeledtraining dataRead newunlabeled dataExtract classesfrom probabilitiesNormalizenumericalfeaturesApply same normalizationas during learningWeak LabelModel Learner Weak Label ModelPredictor Nominal ProbabilityDistribution Creator Joiner Weak Source PMML Reader Decision TreePredictor Joiner Gradient BoostedTrees Learner Gradient BoostedTrees Predictor LogisticRegression Learner Logistic RegressionPredictor Keras NetworkLearner Nominal ProbabilityDistribution Splitter Table Reader Table Reader Keras NetworkExecutor Weak Source Define Network Column Rename(Regex) Normalizer Normalizer (Apply) Binary ClassificationInspector Decision TreePredictor Collect Predictions

Nodes

Extensions

Links

Download

To use this workflow in KNIME, download it from the below URL and open it in KNIME:

Download Workflow
Created by: Adrian Nembach
Created at: 2019-11-22
On NodePit since: 2019-12-05
Last update: 2026-03-24
Created with KNIME version: v4.1.0
Tags: Weak SupervisionWeakly Supervised LearningMachine LearningGradient Boosted TreesLogistic RegressionDeep LearningKeras

Deploy, schedule, execute, and monitor your KNIME workflows locally, in the cloud or on-premises – with our brand new NodePit Runner.

Try NodePit Runner!