cnelias / categoricaltimeseries.jl Goto Github PK

When I try to compute the get_mapping function with values for freq at the edges I get the following error:

julia> mappings = get_mappings(data, f[21])
ERROR: BoundsError: attempt to access 600-element Vector{Float64} at index [-3:45]
Stacktrace:
 [1] throw_boundserror(A::Vector{Float64}, I::Tuple{UnitRange{Int64}})
   @ Base ./abstractarray.jl:651
 [2] checkbounds
   @ ./abstractarray.jl:616 [inlined]
 [3] getindex
   @ ./array.jl:807 [inlined]
 [4] get_mappings(data::Matrix{Any}, freq::Float64; m::Int64)
   @ CategoricalTimeSeries ~/Documents/Programmieren/CategoricalTimeseriesreview/dev/CategoricalTimeSeries/src/SpectralEnvelope.jl:232
 [5] get_mappings(data::Matrix{Any}, freq::Float64)
   @ CategoricalTimeSeries ~/Documents/Programmieren/CategoricalTimeseriesreview/dev/CategoricalTimeSeries/src/SpectralEnvelope.jl:227
 [6] top-level scope
   @ REPL[33]:1

If you would like to restrict it to the boundaries of the se vector you can use max(peak_pos-window,begin):min(peak_pos+window,end) in the indexing.
This would lead to different search window sizes at the edges and I am not sure, whether this would be a problem.

Part of the review for JOSS openjournals/joss-reviews#3733

When I compare the results you get with the Figure 6.4 in C. Weiss I see a similar pattern for the Cohen coefficient, but for the Cramer coefficient we have a different pattern.

I am not sure, whether this is related, In this line of the cramer_coefficient function I suppose you wanted to write += instead of =+

https://github.com/johncwok/CategoricalTimeSeries.jl/blob/01c05c8db6e3127639fecd0f41040046629c5d8e/src/SerialDependence.jl#L160

Otherwise you wouldn't need the for loop.

From the documentation I understand, that the get_mapping function searches for the local maxima near the frequency I specified and gives the same results I would also get in the eigvec vector from the spectral_envelope call, but as a dictionary.
Therefore I assumed, that I would get the same results when I start with the global maximum of the spectral envelope, but for the example data I get the following.
Maybe this is just a misunderstanding on my part.
I would have most of all expected, that the reported strength of peak from the get_mappings function is the same as the maxse value.

julia> maxse, maxpos = findmax(se)
(0.02204087285907974, 401)

julia> get_mappings(data,f[maxpos])
(true_pos, peak_pos) = (399, 401)
position of peak: 0.33 strengh of peak: 0.01
Dict{SubString{String}, Float64} with 4 entries:
  "A" => 0.54
  "T" => -0.57
  "C" => 0.0
  "G" => 0.62

julia> eigvecs[:,maxpos]
3-element Vector{Float64}:
 -0.586861830502818
  0.5964570350305273
  0.5475693538370039

Part of the review for JOSS openjournals/joss-reviews#3733

@JuliaRegistrator register

It would be good to link to the JEREMY BUHLER and MARTIN TOMPA's paper "Finding Motifs Using Random Projections" paper that is mentioned in the motif docs.

Part of the review for JOSS openjournals/joss-reviews#3733

In the example of the correlation functions you use cohen_coefficient for the coefficient computation and then use cramer_coefficient in the bootstrap function.
I think, that this should be the same function.

When I rerun the example in the data clustering documentation, this is not always giving the same result.
This is not stated in the documentation of IB, and was surprising for me.
This is mentioned in the search_optima docstring, but it would be better to also mention this in the IB docstring.
Especially the Bach chorale example confused me, because you state the results as given, but I got different results.
Maybe you could use search_optima in the Bach example, before you start interpreting the results.

As a side note:
I would put the examples further on top of the documentation page, either on top of the further usage or at least on top of the additional functions sections, because examples are the first thing a user searches for.

Part of the review for JOSS openjournals/joss-reviews#3733

In the README you state, that this is regroup of multiple other modules.
Apparently you integrated the code of these modules into this main module and removed the dependency.
Is this correct?
Then you should remove the sentence from the README.

great code
but do you similar project in python ?