Model Monitoring and Debugging
Observe and Explain Your Models
Model Monitoring and Debugging with k-NN+
Howso’s model monitoring and debugging is particularly effective at drift monitoring. This is thanks to our k-NN+ approach and statistical techniques. k-NN+ means that we can add and remove cases rapidly, without expensive re-training operations or creating any number of new models. Additionally, Howso’s statistical measures for analyzing how surprising new data are can be used to detect and thoroughly characterize model drift.
Live Data Connection
Howso can run alongside a traditional ML model and trained on the same training data or a subset of the training data. As data comes in, the surprisal of new data is measured and used to validate if that data is different enough for model drift to be occurring. If this is detected, the model being monitored can be updated along with Howso to ensure that predictive power remains as high as possible.
Best-in-class Debugging
Howso can add its wide array of explanations to predictions made by the ML model. By reacting to production data and the prediction(s) made, Howso can provide highly accurate feature importances, residuals, and other explanations, including hypothetical values and boundary cases, that enable efficient data debugging.
Experience 52% Faster Data Debugging
Howso improves the speed of model monitoring and debugging, saving you time and money.