Data Drift and Concept Drift
Data drift refers to changes in the distribution of input data over time, while concept drift specifically means changes in the relationship between input features and the target variable. Both can degrade the performance of a deployed AI model, as the patterns it learned during training no longer hold true in the current environment. Monitoring for these drifts is crucial for maintaining model performance.
If you train a model to predict fashion trends based on last year's styles, data drift is when people start buying entirely new clothes, and concept drift is when what makes an outfit 'trendy' fundamentally changes.
Understanding and mitigating data and concept drift is vital for the long-term reliability and accuracy of AI systems in real-world, dynamic environments.
Learn one new AI thing every day.
Daily Deck sends you seven plain-English cards like this every morning. Free.
Start free