Training-serving skew and model drift
As decision-makers, it’s important to understand the potential pitfalls in deploying ML models into production. Two of these challenges are training-serving skew and model drift. Let’s explore these concepts, understand their implications, and learn how to mitigate their effects.
Training-serving skew
Training-serving skew occurs when the data used to train a model differs from the data used in serving predictions. This can lead to a significant drop in model performance. Imagine you’re a retail giant, and you’ve trained a model to predict customer purchasing behavior based on historical data. If your model is trained on online sales data but used to predict in-store sales, the skew could lead to inaccurate predictions.
Mitigating training-serving skew
How can we address this? Here are some steps to take:
- Ensure consistency: Make sure that the data used for training and serving is consistent. This...
