AI/ML Assisted labeling in Vertex AI 0

Is there a feature in Vertex AI which will allow AI/ML to assist in labeling data? This usually works by providing a small set of labeled data, followed by a model creation which assists in labeling more data. As more and more data is labeled the model keeps getting better.


Yes, Vertex AI provides a feature called “Active Learning” which enables AI/ML-assisted labeling of data. Active Learning works by training a machine learning model on a small set of labeled data and then using that model to suggest labels for additional data points that are uncertain or ambiguous. This approach allows the model to learn from the labeled data and improve its accuracy as more data is labeled.

To use Active Learning in Vertex AI, you can create an “Active Learning” job and specify the data to be labeled, the machine learning model to use, and the labeling budget (i.e., the number of data points to label). The system will then suggest data points to label based on the model’s uncertainty, and you can review and approve or correct the suggested labels.

Overall, Active Learning can significantly reduce the time and effort required for data labeling, while also improving the accuracy of machine learning models.

Here are some more details on how Active Learning works in Vertex AI:

  1. Initial model training: You start by providing a small set of labeled data to train a machine learning model. This model is used to predict labels for additional data points.
  2. Model prediction: The machine learning model predicts labels for a batch of unlabeled data points, and the system selects the data points for which the model is most uncertain or ambiguous. These are the data points that will be suggested for labeling.
  3. Labeling: The suggested data points are presented to a human labeler, who reviews the suggested label and either approves or corrects it. The labeled data is then added to the training set for the machine learning model.
  4. Model retraining: The machine learning model is retrained on the newly labeled data, and the process repeats from step 2
  5. This iterative process continues until the desired level of accuracy is achieved or the labeling budget is exhausted.

It’s worth noting that Active Learning can be used with various types of machine learning models, including both supervised and unsupervised models. Additionally, Vertex AI provides a range of tools and APIs for data labeling, including human-in-the-loop labeling, where machine-learning models assist human labelers in the labeling process.