(Also found in highly technical)
Hi,
I'm interested in using CTW's in a project, but am a little bit in doubt about a couple of things. I haven't been able to find a good full description of it, though I've managed to piece together some info, but I'm not quite clear on everything (even the algorithm).
Is the CTW's depth fixed or just very slowly growing? I understand that this is supposed to a major advantage vis-a-vis Lempel-Ziv, but if the depth is fixed, how does one determine the appropriate depth for a tree (other than trial and error)? I'm also a bit confused about the compression, considering that it's based off of a full binary tree... if it uses something analoguos to the Lempel-Ziv dictionary, then the compression definitely shouldn't better (i.e. representing a leaf costs just as much as the context of that leaf).
Does decoding require the tree, or is there some analogous structure to the Lempel-Ziv dictionary? How does the information get decoded?
There's quite a lot about parameters, and when not knowing them to use the Zero-Redundancy Estimator, instead of the Krichevsky-Trofimov Estimator. What paramaters are there, and how should one decide which of the estimators to use?
Thanks.
Hi,
I'm interested in using CTW's in a project, but am a little bit in doubt about a couple of things. I haven't been able to find a good full description of it, though I've managed to piece together some info, but I'm not quite clear on everything (even the algorithm).
Is the CTW's depth fixed or just very slowly growing? I understand that this is supposed to a major advantage vis-a-vis Lempel-Ziv, but if the depth is fixed, how does one determine the appropriate depth for a tree (other than trial and error)? I'm also a bit confused about the compression, considering that it's based off of a full binary tree... if it uses something analoguos to the Lempel-Ziv dictionary, then the compression definitely shouldn't better (i.e. representing a leaf costs just as much as the context of that leaf).
Does decoding require the tree, or is there some analogous structure to the Lempel-Ziv dictionary? How does the information get decoded?
There's quite a lot about parameters, and when not knowing them to use the Zero-Redundancy Estimator, instead of the Krichevsky-Trofimov Estimator. What paramaters are there, and how should one decide which of the estimators to use?
Thanks.
Facebook
Twitter