Skip to main content

Documentation Index

Fetch the complete documentation index at: https://arize-ax.mintlify.dev/docs/llms.txt

Use this file to discover all available pages before exploring further.

​
What’s New

March 28, 2022 Free Arize Account! The free offering makes it easy for ML engineers to get up and running in minutes so that you can detect, root cause, and resolve model performance issues faster. Featuring an easy integration via an SDK or file ingestion from major cloud storage providers, ML teams can begin monitoring, troubleshooting, and improving model performance. Begin your ML Observability journey and sign up for your free account today! Spaces Arize accounts now consist of organizations and spaces to support larger teams and protect work across multiple business units.
  • Organizations consist of one or more *Spaces. *They represent a single business unit with a common purpose/ function and can make up several collaborating teams.
  • Spaces represent an environment for groups of similar models within an organization. They can be used as a safe experimental environment or to promote collaboration across models.
  • At each level permissions are protected by Role-Based Access Control (RBAC), allowing customers to create isolated and protected environments across their business
Quickstart Guide Any new user of Arize will be greeted with a quick start guide that includes:
  • Walkthrough videos of the Arize platform & its capabilities
  • Step by Step tutorials of uploading a test model data via Python Batch (arize.pandas), Python Realtime (arize.log) + Java & Cloud Storage Ingestion Methods
  • A fraud demo model overview and Colab notebook walkthrough for sending a fraud model example to Arize
Access the quick start guide at any time from the bottom left of the navigation bar.

​
Enhancements

March 14, 2022 Python SDK 3.4.0 Adds surrogate_explainability= flag to the pandas logger. Using the surrogate explainability approach, users have the option to pass a flag with a request to send data that would produce SHAP values. When the flag is enabled, a tree-based surrogate model is trained using the dataset’s features and predictions. The surrogate model then generates SHAP values before sending the combined dataset to the Arize platform. With surrogate explainability, users can now easily generate feature importance values without having to maintain an extra computation pipeline. Learn more about surrogate models here.
Delete File Import March 28, 2022 This allows the option to easily delete import jobs from your cloud storage to improve overall data pipeline workflows. Default Baseline for New Production Models New models added will have a default baseline defined by the last 30 days of production data. This will enable the Arize platform to auto-populate model health metrics on the Model Overview page given 30 days of production data sent to the platform. To read more about baselines, you can check out our docs here. **Python SDK **4.0.0 Allows users to use the new Organizations/Spaces organizational hierarchy.
  • Authentication request change fromorganization_keyto space_key
Model Overview Tab Improvements
  • Addition of a ‘datasets’ card to view the latest datasets added to the platform at a glance
  • “Jump to Latest Data” button to easily navigate to the latest data in your dataset
Dimension Details Tab The drift tab now has the ability to filter by more specific features, prediction values, tags, or actuals from the toolbar for a more granular analysis of drift.

​
In the News

March 14, 2022 Arize:Observe - All Day Virtual Event, March 29th Tune into Arize:Observe on March 29th to hear from industry experts like UMAP’s Founder Leland McInnes, Coinbase Director of Engineering Chintan Turakhia, Chick-fil-A Senior Lead Machine Learning Engineer Korri Jones, MBA, Uber Director of Engineering Smitha Shyam, and more! Register now!
Rise of the ML Engineer: Flávio Clésio, Artsy In a wide-ranging interview, ML engineer Flávio Clésio of Artsy stresses the importance of model monitoring and explains why it’s all about “exploit and explore” with recommendation systems. Read more.
Your Data Science Workflows Are About To Get A Lot More Scalable March 28, 2022 What would enterprise-ready Pandas mean for data scientists? We recently caught up with Ponder co-founder and CEO Doris Lee a week after the company’s $7 million seed round to find out in this wide-ranging interview. Read it.