2 ms·

I'd be concerned that memoryless is misleading for markovian. A markov chain can be used to represent a fixed, nonzero memory, since if you have an n-state dfa

I'd be concerned that memoryless is misleading for markovian. A markov chain can be used to represent a fixed, nonzero memory, since if you have an n-state dfa D you can take the set of pairs of the states of D (or specifically D + null) and reweight them. The markov chain over D^2 has no memory over D^2, but does have memory over D, and this is often a useful property (bigrams). You repeat this operation to get more and more (but always a fixed constant maximum) history.

mehrdadn 6y ago
Memoryless isn't even something I came up with, it's practically in the first sentence of everyone's description of the Markov property, e.g.: https://en.wikipedia.org/wiki/Markov_chain#Definition https://en.wikipedia.org/wiki/Markov_chain#Definition Also see: https://en.wikipedia.org/wiki/Memorylessness https://en.wikipedia.org/wiki/Memorylessness ("memorylessness refers to the Markov property") All I'm doing is suggesting just using the existing description as the name itself. Even if there were some cases where it wouldn't apply, though, it wouldn't really matter. Fourier domain also applies to things other than time but that doesn't make frequency domain a bad term. It's just a name meant to carry some intuition; that's all. It's not meant to serve as the formal definition.