This release includes several stability improvements, including updates to the baseline model metrics, data profiling view, and data labeling workflow.

Platform

1. A Personalized Onboarding

   We have personalized the onboarding flow to guide you.

Data Profiling Updates

1. Updated Data Profiling Page

   We have added new widgets to get more insights about your registered datasets. You can now see four more charts for classification datasets.

   1. The class-wise Markov Quality Score for the labeled dataset.

      This provides a quantitative measure of potential mislabels at the class level in a given dataset. Markov Quality Score tasklet should be run to see this score.

   2. Top k features for Tabular datasets

      Identifies top-K features with more predictive information to identify the target. Markov Quality Score tasklet should be run to see this score.

   3. Baseline Model Metrics

      Provides classification metrics such as precision, recall, f1-score and accuracy for the baseline model. The baseline Model Tasklet should be run to see these metrics and create a baseline model app.

   4. Overlapping classes

      This widget shows what percentage of instances labeled as one class in the original dataset were more likely to be classified as another.

Relabeling Flow Updates

1. Run Analysis On Relabeled Dataset
   Users can now select and run analyzers on the corrected dataset after completing any relabeling flow.

Bug fixes

1. Several stability, UX, and other bug fixes have been addressed across the product.

📢 MarkovML's Community Spotlight

Join the conversation at markovml.com and discover how our community of 300+ domain experts discusses everything in ML.