up:: [[Machine Learning MOC]] ^qbglai # Supervised Machine Learning Data Exists in a Euclidean Space All training data exists as points in a multidimensional Euclidean space. You can visualize an enclosure of this data (the *training domain*). Within that space, supervised ML acts as interpolation geometry. ## Visualization ![[Pasted image 20220324115913.png]] In reality, the training domain will look jagged and weird because data is messy, incomplete, and often doesn't line up with the regional problem space. Because ML models can not extrapolate, the model won't work well at the edges or outside of the training domain. ## Addressing Gaps Between the Training Domain and Problem Space **Active Learning** can address these gaps. This is the process of flagging confusing inputs during inference and setting them aside for human ground-truthing. Once those inputs are correctly labeled, they're added to the training data to fill gaps in the training domain. ^dd9a2d ## Source ![[Sources/Videos/FourthBrain - ML Model Drift and Decay.md#^jea50m]] ![[Sources/Videos/FourthBrain - ML Model Drift and Decay.md#^hhdfps]] ![[Sources/Videos/FourthBrain - ML Model Drift and Decay.md#^tkrs18]]