A Recursive Model Index (RMI) uses a hierarchy of models to…

A Recursive Model Index (RMI) uses a hierarchy of models to approximate the mapping from a key to its position in a sorted array. Consider a simple two-level RMI: • A single top-level model that maps a key to a coarse predicted position or to a bucket ID. • Many bottom-level models, each trained on a small partition of keys with nearly linear behavior.You can choose between two model families: • Neural network (NN) • Linear regression (LR)Which combination of models at the top and bottom levels is generally the most sensible choice for balancing accuracy and lookup latency in a learned index?

A student claims: “An R-tree is basically a B+Tree for recta…

A student claims: “An R-tree is basically a B+Tree for rectangles. For a point lookup, we follow exactly one path from the root to a leaf, just like in a B+Tree, because at each level we can choose the single child whose bounding box contains the point.”Which statement about search behavior in R-trees versus B+Trees is NOT correct?

A column store holds three columns, each originally stored a…

A column store holds three columns, each originally stored as 8-byte values per row, except the string column which averages 20 bytes per value: a: 64-bit sorted integers, delta-encoded with variable-length coding so that the average compressed size is 2 bytes per value. b: 64-bit floats, quantized lossily to 32-bit floats. c: variable-length strings averaging 20 bytes per value, but with only 200 distinct values; c is dictionary-encoded.Assuming all three columns have the same number of rows, which statement about their compression ratios is most accurate?