Active learning

Active learning is a specific approach in machine learning where a a model can achieve better performance with a lower amount of training instances (see Figure below).

Figure 1

A quick presentation of the results can be shown in Figure 2. However for a more detailed explanation of the equations and content, can be found in the notebook Active learning.ipynb

Figure 2: (a) is a toy data set. (b) Logistic regression is trained on randomly sampled instances (downward triangles). (c) Logistic regression is trained on "actively learned" instances (downward triangles), creating a more accurate decision boundary