Kristian Agasøster Haaga
Kristian Agasøster Haaga
> Unfortunately, I was too optimistic with that. I have a rough draft ready, but unfortunately, I can't force the inverse non-uniform fourier transform to return a real valued time...
## The issue of subjectivity > Hypothesis testing needs discriminatory statistics. Choosing a discriminatory statistic is a subjective thing, not an objective one. There is no mathematical equation that tells...
Here's an example of how I do it in CausalityTools.jl. There are 3 state variables, each depending on states a maximum of 2 time steps back. Hence, `past_states` is an...
> Isn't this an N-state markov chain? Or, that's the most general way to define these systems? That's what we should implement probably From the top of my head, I...
Another example which is deterministic + stochastic: ```julia Base.@kwdef struct Logistic2Unidir{V, C, R1, R2, Σy, R}
Since this involves averaging multiple periodograms, I guess it will be more time-consuming than using a single estimate of the periodogram? I suggest that if you implement this, @duodenum96, the...
Excellent! Looking forwards to your contribution :D
mutualinfo with KSG1/KSG2's results different from scikit-learn and about 2x slower for large arrays
Hey, @liuyxpp! Thanks for the report. Can you report back the output of `Pkg.status()` in the same session as you're testing, so I know which versions of packages you're using?...
mutualinfo with KSG1/KSG2's results different from scikit-learn and about 2x slower for large arrays
@liuyxpp Also: are you running also sort of multithreading for the Python code?
mutualinfo with KSG1/KSG2's results different from scikit-learn and about 2x slower for large arrays
It would also be nice to see whether this apparent performance difference is also true for higher dimensional data. In the Ross paper which is linked in the python docs,...