Weak Label Model Learner

Learns a generative label model from the provided label source columns. This node is a key component for the realization of weak supervision approaches as popularized by Snorkel . The idea in weak supervision is that it is often possible to create a number of simple inaccurate models (e.g. simple rules or existing models for slightly different tasks) that can label unlabeled data and that the agreements and disagreements of these simple models can be analyzed to infer information on the true label. Our implementation is a TensorFlow based adaptation of the matrix completion approach proposed in this paper by the Snorkel team. We refer to the publication for details on the strategy.

Options

Label sources

Select the columns which act as label sources i.e. that contain noisy labels for some of the rows in the first input table. It is assumed that a missing value means that the respective label source did not label the corresponding row.

Epochs

The number of optimization steps to perform. More epochs can result in better results but also directly translate into a longer runtime.

Learning rate

The learning rate dictates how much a single training epoch changes the learned model. A smaller learning rate requires more epochs to reach convergence while a large learning rate might lead to divergence of the algorithm.

Views

This node has no views

Workflows

• 01_Guided_Labeling_for_Document_ClassificationKNIME Hub
• 01_Guided_Labeling_for_Document_Classification-FIXED-HEADERKNIME Forum
• 03_Weak_Supervision_for_Document_ClassificationKNIME Hub
• 05_Weak_Supervision_on_the_Adult_datasetKNIME Hub

Links

• No links available

Developers