// MODEL OPTIMIZATION AND PROMPT SYNTAX TERM
Data Drift
Data drift occurs when the characteristics of the data used for predictions change over time, making the model less accurate.
TECHNICAL DEFINITION
Data drift occurs when the statistical properties of the input data to a machine learning model change over time in an unforeseen way, leading to a degradation in model performance because the model was trained on different data distributions.
BACKGROUND
Adversarial machine learning is the study of the attacks on machine learning algorithms, and of the defenses against such attacks.
READ MORE ON WIKIPEDIASYNONYMS & ALIASES
- Input data drift
- covariate shift
- feature drift
USAGE NOTE
Detecting data drift often triggers model retraining or investigation.
DEVELOPERS
Organizations developing technology related to Data Drift.
Arize AI provides an AI Observability Platform that helps machine learning teams monitor, troubleshoot, and explain models in production, with a core focus on detecting and analyzing data and model drift.
WhyLabs offers an AI Observability Platform called WhyLabs AI Observatory, which provides continuous monitoring for AI applications to detect data drift, concept drift, and data quality issues.
Fiddler AI provides an AI Observability platform that helps enterprises monitor, explain, and improve their AI models, including robust capabilities for detecting and analyzing data and model drift.
Weights & Biases (W&B) is an MLOps platform that offers tools for experiment tracking, model versioning, and model monitoring, including features to detect data and concept drift in production models.
Seldon provides an open-source and enterprise MLOps platform for deploying, managing, and monitoring machine learning models at scale, featuring robust drift detection capabilities to ensure model reliability.
Amazon SageMaker Model Monitor is a service within AWS that continuously monitors the quality of machine learning models in production, automatically detecting data drift and concept drift.
Microsoft Azure Machine Learning offers comprehensive MLOps capabilities, including tools for model monitoring, data drift detection, and maintaining the performance of deployed AI models.
Datadog, a leading monitoring and analytics platform, extends its capabilities to ML monitoring, allowing users to track model performance and detect data drift as part of their observability offerings.