improved

Product Release: August 23, 2023

We have made several enhancements and introduced new features to our MarkovML. Here are some of the updates.

Platform Updates

  1. Email Signup

We have recently included the option to sign up using your email and the existing Google sign-up. This provides another easy option to create a MarkovML account and simplifies onboarding.

Redesigned sign-up page

Redesigned sign-up page

  1. Sample data for all workspaces

We've included sample datasets, experiments, and evaluations to help new users explore the platform and understand its capabilities.

A sample dataset page showcasing everything MarkovML has to offer for datasets

A sample dataset page showcasing everything MarkovML has to offer for datasets

Snippets

  1. Embeddings in snippets

MarkovML snippets are particularly powerful because they allow for directly embedding charts from experiments, evaluations, and datasets. Our users use Snippets to create interactive reports and insights with their teams. We are pleased to introduce support for adding an embedding view to snippets πŸŽ‰

Add embeddings in Snippets and interact with the chart right here

Add embeddings in Snippets and interact with the chart right here

Notebooks

  1. Faster Notebooks

Notebooks offer a convenient and easy-to-use solution for a hosted JupyterHub experience. With MarkovSDK installed and popular machine-learning libraries, you can get started quickly. Our average load times for starting the notebook have been reduced to under 30 seconds, making the experience much smoother.πŸš€

Quicker load time of about 30 seconds for Notebooks

Quicker load time of about 30 seconds for Notebooks

  1. Notebook with GPU support (beta)

We have included the option to set up your MarkovML Notebook server with GPU and CPU configurations. Note that this feature is currently in beta and will only be available upon request after review. To request GPU support, kindly send an email to [email protected]

GPU support for Notebooks in Large Machine

GPU support for Notebooks in Large Machine

Datasets

  1. Introducing Dataset Label Correction Workflow

We have incorporated a new Mislabel correction feature into MarkovML. With this new addition, users can now easily rectify the possibly mislabeled points detected by the Markov Quality Score Analysis within the platform and create a new corrected version.

πŸ“˜

This feature is coming to enterprise customers in a few days time. Our team will reach out to you with updates.

Introducing Relabelling workflow of mis-labelled dataset

Introducing Relabelling workflow of the mislabelled dataset

  1. Support for Unlabeled & Continous target datasets

MarkovML now supports the analysis of unlabeled datasets. During the dataset registration process, select This is an unlabelled dataset option to register an unlabeled dataset.

You can also add datasets with continuous target values, for example, regression datasets for analysis.

  1. On-demand Analysis on Dataset Registration

Users can now select which analyzers to run during the registration step. Only run what you need to save time & cost.

Bug Fixes: We have also made minor improvements in our dataset registration and onboarding flow. A lot of bugs have also been quashed. See you later with more updates!