- Skimmed this, but just making note to self
- Does form of Lempel-Ziv compression on continuous valued time series
- In particular, its for finding similarity between time series, like methods of dynamic time warping
- Works on continuous data by computing a distance <perhaps Euclidian?> of each point in the time – if its epsilon close to something else in the dictionary, it is called that.
- This allows for a discretization, but one thats not based on coarse quantization
- They claim this method works as well as other methods for finding similarity between time series but this approach is much cheaper.