Compression of floating point timeseries
I recently had cause to investigate fast methods of storing and transferring financial timeseries. Naively, timeseries can be represented in memory or on disk as simple dense arrays of floating point numbers. This is an attractive representation with many nice properties: