Learned Spatial Keyword Index
My final year thesis project at HKUST was about proposing, implementing and testing learned spatial keyword indexes.
The original paper that introduced the concept of a learned index is from ,
The B-Tree is the most fundamental index structure, supporting 1D range indexes.
The B-Tree can be generalized into a model, with inputs being the element's key, and the output being the element's position in the array.
This essentially means we are learning the CDF of array.
The rationale behind learned indexes is then to use machine learning techniques to
There are many nuances to this application of machine learning. For example, unlike traditional ML tasks, we want the model to overfit as much as possible, as it will be able to better
The goal of my project is to implement and test learned spatial keyword indexes.