// MODEL OPTIMIZATION AND PROMPT SYNTAX TERM

Model Drift

Model drift happens when a deployed model's performance gets worse over time because the real-world data it sees has changed.

TECHNICAL DEFINITION

Model drift, also known as performance drift, refers to the degradation of a machine learning model's predictive accuracy or effectiveness in a production environment due to changes in the underlying data distribution or relationships between features and targets.

BACKGROUND

In machine learning, diffusion models, also known as diffusion-based generative models or score-based generative models, are a class of latent variable generative models. A diffusion model consists of two major components: the forward diffusion process, and the reverse sampling process. The goal of diffusion models is to learn a diffusion process for a given dataset, such that the process can generate new elements that are distributed similarly as the original dataset. A diffusion model models data as generated by a diffusion process, whereby a new datum performs a random walk with drift through the space of all possible data. A trained diffusion model can be sampled in many ways, with different efficiency and quality.

READ MORE ON WIKIPEDIA

SYNONYMS & ALIASES

  • Performance drift
  • model decay
  • accuracy degradation

USAGE NOTE

Model drift necessitates retraining or updating the deployed model.

DEVELOPERS

Organizations developing technology related to Model Drift.

  • Arize AI

    Specializes in AI observability, offering tools for detecting and diagnosing model drift, data quality issues, and performance degradation in production machine learning models.

  • Fiddler AI

    Provides an AI observability platform for monitoring, explainability, and fairness, with robust capabilities to detect and analyze model drift, data drift, and data integrity issues.

  • Arthur AI

    Offers an AI performance monitoring platform that helps detect and diagnose model drift, data quality issues, and other performance problems for production AI systems, ensuring model reliability.

  • Google Cloud

    Their Vertex AI platform includes MLOps capabilities for monitoring machine learning models in production, with features specifically designed to detect and alert on model drift and data drift.

  • Microsoft Azure

    Azure Machine Learning offers comprehensive MLOps tools, including model monitoring services that help detect drift in model predictions and input data, crucial for maintaining model performance.

  • Amazon Web Services (AWS)

    AWS SageMaker Model Monitor provides continuous monitoring of ML models in production to detect data and model quality issues, including drift, and offers explainability for identified issues.

  • Weights & Biases

    An MLOps platform used for experiment tracking, model versioning, and production monitoring, offering dashboards and alerts for identifying performance issues and drift in machine learning models.

  • Seldon

    Specializes in MLOps, providing enterprise-grade solutions for deploying, managing, and monitoring machine learning models in production, including advanced drift detection capabilities.

RELATED TERMS IN MLOPS & DEPLOYMENT