Format

src/CovarianceImage.jl

@@ -9,7 +9,7 @@ function clustercovariance(Σ²)
     if !issymmetric(Dr)
         @warn "Correlation distance matrix is not symmetric, so not clustering"
     end
-    Clustering.hclust(Dr; linkage=:average, branchorder=:optimal)
+    Clustering.hclust(Dr; linkage = :average, branchorder = :optimal)
 end
 
 """
@@ -27,7 +27,10 @@ Either provide as positional arguments a vector `f` of N row names and an N×_ m
 """
 covarianceimage
 @userplot CovarianceImage
-Plots.@recipe function f(g::CovarianceImage; palette=[:cornflowerblue, :crimson, :forestgreen], colormode=:top, colorbargrad=:binary, donames=true, docluster=true, verbose=true, dendrogram=false)
+Plots.@recipe function f(g::CovarianceImage;
+                         palette = [:cornflowerblue, :crimson, :forestgreen],
+                         colormode = :top, colorbargrad = :binary, donames = true,
+                         docluster = true, verbose = true, dendrogram = false)
     if g.args[1] isa AbstractFeatureArray || g.args[1] isa AbstractDimArray
         f, Σ² = string.(getdim(g.args[1], 1)), g.args[1] |> Array
     elseif length(g.args) == 2 && g.args[2] isa AbstractMatrix
@@ -38,11 +41,12 @@ Plots.@recipe function f(g::CovarianceImage; palette=[:cornflowerblue, :crimson,
 
     issymmetric(Σ²) || (Σ² = cov(Σ²'))
 
-    any(diag(Σ²) .== 0) && @warn "Covariance matrix is not positive definite, which may cause an error"
+    any(diag(Σ²) .== 0) &&
+        @warn "Covariance matrix is not positive definite, which may cause an error"
     linecolor --> nothing
     if docluster == true
         idxs = clustercovariance(Σ²).order
-    elseif docluster isa Union{AbstractVector,Tuple}
+    elseif docluster isa Union{AbstractVector, Tuple}
         idxs = docluster # Precomputed indices
     elseif docluster isa Clustering.Hclust
         idxs = docluster.order
@@ -58,11 +62,14 @@ Plots.@recipe function f(g::CovarianceImage; palette=[:cornflowerblue, :crimson,
         colorbar --> true
     else
         λ = (eigvals ∘ Symmetric ∘ Array)(Σ̂²)
-        λi = sortperm(abs.(λ), rev=true)
+        λi = sortperm(abs.(λ), rev = true)
         λ = λ[λi]
         P = (eigvecs ∘ Symmetric ∘ Array)(Σ̂²)[:, λi] # Now sorted by decreasing eigenvalue norm
-        vidxs = sortperm(abs.(P[:, 1]), rev=true)
-        verbose && isnothing(printstyled("Feature weights:\n", color=:red, bold=true)) && display(vcat(hcat("Feature", ["PC$i" for i ∈ 1:N]...), hcat(f̂[vidxs], round.(P[vidxs, 1:N], sigdigits=3))))
+        vidxs = sortperm(abs.(P[:, 1]), rev = true)
+        verbose &&
+            isnothing(printstyled("Feature weights:\n", color = :red, bold = true)) &&
+            display(vcat(hcat("Feature", ["PC$i" for i in 1:N]...),
+                         hcat(f̂[vidxs], round.(P[vidxs, 1:N], sigdigits = 3))))
         P = abs.(P)
         # if colormode isa Matrix # Supply custom coloring matrix. Should be an nfeature×ncolor matrix
         #     P = colormode
@@ -71,30 +78,30 @@ Plots.@recipe function f(g::CovarianceImage; palette=[:cornflowerblue, :crimson,
         #     𝑓′ = parse.(Colors.XYZ, palette[1:N]);
         if colormode == :top # * Color by the number of PC's given by the length of the color palette
             P = P[:, 1:N]
-            P̂ = P .^ 2.0 ./ sum(P .^ 2.0, dims=2)
+            P̂ = P .^ 2.0 ./ sum(P .^ 2.0, dims = 2)
             # Square the loadings, since they are added in quadrature. Maybe not a completely faithful representation of the PC proportions, but should get the job done.
             𝑓′ = parse.(Colors.XYZ, palette[1:N])
         elseif colormode == :all # * Color by all PC's. This can end up very brown
             Σ̂′² = Diagonal(abs.(λ))
-            P̂ = P .^ 2.0 ./ sum(P .^ 2.0, dims=2)
+            P̂ = P .^ 2.0 ./ sum(P .^ 2.0, dims = 2)
             p = fill(:black, size(P, 2))
             p[1:N] = palette[1:N]
             𝑓′ = parse.(Colors.XYZ, p)
-            [𝑓′[i] = Σ̂′²[i, i] * 𝑓′[i] for i ∈ 1:length(𝑓′)]
+            [𝑓′[i] = Σ̂′²[i, i] * 𝑓′[i] for i in 1:length(𝑓′)]
         end
         𝑓 = Vector{eltype(𝑓′)}(undef, size(P̂, 1))
         try # Load colors by PC weights
             𝑓 = P̂ * 𝑓′
         catch
             # Equivalent but slower
             @info "Iterating to load covariances"
-            for ii ∈ 1:length(𝑓)
-                𝑓[ii] = sum([P̂[ii, jj] * 𝑓′[jj] for jj ∈ 1:length(𝑓′)])
+            for ii in 1:length(𝑓)
+                𝑓[ii] = sum([P̂[ii, jj] * 𝑓′[jj] for jj in 1:length(𝑓′)])
             end
         end
 
         H = Array{Colors.XYZA}(undef, size(Σ̂²))
-        for (i, j) ∈ Tuple.(CartesianIndices(H)) # Apply the correlations as transparencies
+        for (i, j) in Tuple.(CartesianIndices(H)) # Apply the correlations as transparencies
             J = (𝑓[i] + 𝑓[j]) / 2
             H[i, j] = Colors.XYZA(J.x, J.y, J.z, A[i, j])
         end
@@ -106,7 +113,7 @@ Plots.@recipe function f(g::CovarianceImage; palette=[:cornflowerblue, :crimson,
             seriescolor --> colorbargrad
         end
         if backend() == Plots.GRBackend() # For some reason, GR does heatmaps differently
-            xs = 0.5:1:size(H, 1)+0.5
+            xs = 0.5:1:(size(H, 1) + 0.5)
         else
             xs = 1:size(H, 1)
         end
@@ -127,7 +134,7 @@ Plots.@recipe function f(g::CovarianceImage; palette=[:cornflowerblue, :crimson,
         markercolor := colorbargrad
         (zeros(2), zeros(2))
     end
-    for i ∈ 1:N
+    for i in 1:N
         @series begin
             seriestype := :shape
             if colormode != :raw

src/metadata.jl

@@ -1,26 +1,25 @@
-const featurenames = [
-    :DN_HistogramMode_5
-    :DN_HistogramMode_10
-    :CO_Embed2_Dist_tau_d_expfit_meandiff
-    :CO_f1ecac
-    :CO_FirstMin_ac
-    :CO_HistogramAMI_even_2_5
-    :CO_trev_1_num
-    :DN_OutlierInclude_p_001_mdrmd
-    :DN_OutlierInclude_n_001_mdrmd
-    :FC_LocalSimple_mean1_tauresrat
-    :FC_LocalSimple_mean3_stderr
-    :IN_AutoMutualInfoStats_40_gaussian_fmmi
-    :MD_hrv_classic_pnn40
-    :SB_BinaryStats_diff_longstretch0
-    :SB_BinaryStats_mean_longstretch1
-    :SB_MotifThree_quantile_hh
-    :SC_FluctAnal_2_rsrangefit_50_1_logi_prop_r1
-    :SC_FluctAnal_2_dfa_50_1_2_logi_prop_r1
-    :SP_Summaries_welch_rect_area_5_1
-    :SP_Summaries_welch_rect_centroid
-    :SB_TransitionMatrix_3ac_sumdiagcov
-    :PD_PeriodicityWang_th0_01]
+const featurenames = [:DN_HistogramMode_5
+                      :DN_HistogramMode_10
+                      :CO_Embed2_Dist_tau_d_expfit_meandiff
+                      :CO_f1ecac
+                      :CO_FirstMin_ac
+                      :CO_HistogramAMI_even_2_5
+                      :CO_trev_1_num
+                      :DN_OutlierInclude_p_001_mdrmd
+                      :DN_OutlierInclude_n_001_mdrmd
+                      :FC_LocalSimple_mean1_tauresrat
+                      :FC_LocalSimple_mean3_stderr
+                      :IN_AutoMutualInfoStats_40_gaussian_fmmi
+                      :MD_hrv_classic_pnn40
+                      :SB_BinaryStats_diff_longstretch0
+                      :SB_BinaryStats_mean_longstretch1
+                      :SB_MotifThree_quantile_hh
+                      :SC_FluctAnal_2_rsrangefit_50_1_logi_prop_r1
+                      :SC_FluctAnal_2_dfa_50_1_2_logi_prop_r1
+                      :SP_Summaries_welch_rect_area_5_1
+                      :SP_Summaries_welch_rect_centroid
+                      :SB_TransitionMatrix_3ac_sumdiagcov
+                      :PD_PeriodicityWang_th0_01]
 
 const catch24_featurenames = [featurenames..., :DN_Mean, :DN_Spread_Std]
 
@@ -51,51 +50,50 @@ const short_featurenames = [
 const short_catch24_featurenames = [short_featurenames..., :mean, :std]
 
 const featuretypes = Dict(featurenames .=> [
-    #DN_HistogramMode_5
-    Cdouble
-    #DN_HistogramMode_10
-    Cdouble
-    #CO_Embed2_Dist_tau_d_expfit_meandiff
-    Cdouble
-    #CO_f1ecac
-    Cdouble
-    #CO_FirstMin_ac
-    Cint
-    #CO_HistogramAMI_even_2_5
-    Cdouble
-    #CO_trev_1_num
-    Cdouble
-    #DN_OutlierInclude_p_001_mdrmd
-    Cdouble
-    #DN_OutlierInclude_n_001_mdrmd
-    Cdouble
-    #FC_LocalSimple_mean1_tauresrat
-    Cdouble
-    #FC_LocalSimple_mean3_stderr
-    Cdouble
-    #IN_AutoMutualInfoStats_40_gaussian_fmmi
-    Cdouble
-    #MD_hrv_classic_pnn40
-    Cdouble
-    #SB_BinaryStats_diff_longstretch0
-    Cdouble
-    #SB_BinaryStats_mean_longstretch1
-    Cdouble
-    #SB_MotifThree_quantile_hh
-    Cdouble
-    #SC_FluctAnal_2_rsrangefit_50_1_logi_prop_r1
-    Cdouble
-    #SC_FluctAnal_2_dfa_50_1_2_logi_prop_r1
-    Cdouble
-    #SP_Summaries_welch_rect_area_5_1
-    Cdouble
-    #SP_Summaries_welch_rect_centroid
-    Cdouble
-    #SB_TransitionMatrix_3ac_sumdiagcov
-    Cdouble
-    #PD_PeriodicityWang_th0_01
-    Cint])
-
+                           #DN_HistogramMode_5
+                           Cdouble
+                           #DN_HistogramMode_10
+                           Cdouble
+                           #CO_Embed2_Dist_tau_d_expfit_meandiff
+                           Cdouble
+                           #CO_f1ecac
+                           Cdouble
+                           #CO_FirstMin_ac
+                           Cint
+                           #CO_HistogramAMI_even_2_5
+                           Cdouble
+                           #CO_trev_1_num
+                           Cdouble
+                           #DN_OutlierInclude_p_001_mdrmd
+                           Cdouble
+                           #DN_OutlierInclude_n_001_mdrmd
+                           Cdouble
+                           #FC_LocalSimple_mean1_tauresrat
+                           Cdouble
+                           #FC_LocalSimple_mean3_stderr
+                           Cdouble
+                           #IN_AutoMutualInfoStats_40_gaussian_fmmi
+                           Cdouble
+                           #MD_hrv_classic_pnn40
+                           Cdouble
+                           #SB_BinaryStats_diff_longstretch0
+                           Cdouble
+                           #SB_BinaryStats_mean_longstretch1
+                           Cdouble
+                           #SB_MotifThree_quantile_hh
+                           Cdouble
+                           #SC_FluctAnal_2_rsrangefit_50_1_logi_prop_r1
+                           Cdouble
+                           #SC_FluctAnal_2_dfa_50_1_2_logi_prop_r1
+                           Cdouble
+                           #SP_Summaries_welch_rect_area_5_1
+                           Cdouble
+                           #SP_Summaries_welch_rect_centroid
+                           Cdouble
+                           #SB_TransitionMatrix_3ac_sumdiagcov
+                           Cdouble
+                           #PD_PeriodicityWang_th0_01
+                           Cint])
 
 """
     Catch22.featurekeywords
@@ -147,53 +145,52 @@ const featurekeywords = [ # See hctsa
     #PD_PeriodicityWang_th0_01
     ["periodicity", "spline"]]
 
-
 """
     Catch22.featuredescriptions
 A vector listing short descriptions of each feature, as strings.
 """
 const featuredescriptions = [  # See catch22 paper
-    #DN_HistogramMode_5
-    "Mode of z-scored distribution (5-bin histogram)"
-    #DN_HistogramMode_10
-    "Mode of z-scored distribution (10-bin histogram)"
-    #CO_Embed2_Dist_tau_d_expfit_meandiff
-    "Exponential fit to successive distances in 2-d embedding space"
-    #CO_f1ecac
-    "First 1/𝑒 crossing of autocorrelation function"
-    #CO_FirstMin_ac
-    "First minimum of autocorrelation function"
-    #CO_HistogramAMI_even_2_5
-    "Automutual information, 𝑚=2,𝜏=5"
-    #CO_trev_1_num
-    "Time-reversibility statistic, ⟨(𝑥ₜ₊₁−𝑥ₜ)³⟩ₜ"
-    #DN_OutlierInclude_p_001_mdrmd
-    "Time intervals between successive extreme events above the mean"
-    #DN_OutlierInclude_n_001_mdrmd
-    "Time intervals between successive extreme events below the mean"
-    #FC_LocalSimple_mean1_tauresrat
-    "Change in correlation length after iterative differencing"
-    #FC_LocalSimple_mean3_stderr
-    "Mean error from a rolling 3-sample mean forecasting"
-    #IN_AutoMutualInfoStats_40_gaussian_fmmi
-    "First minimum of the automutual information function"
-    #MD_hrv_classic_pnn40
-    "Proportion of successive differences exceeding 0.04𝜎 (Mietus 2002)"
-    #SB_BinaryStats_diff_longstretch0
-    "Longest period of successive incremental decreases"
-    #SB_BinaryStats_mean_longstretch1
-    "Longest period of consecutive values above the mean"
-    #SB_MotifThree_quantile_hh
-    "Shannon entropy of two successive letters in equiprobable 3-letter symbolization"
-    #SC_FluctAnal_2_rsrangefit_50_1_logi_prop_r1
-    "Proportion of slower timescale fluctuations that scale with linearly rescaled range fits"
-    #SC_FluctAnal_2_dfa_50_1_2_logi_prop_r1
-    "Proportion of slower timescale fluctuations that scale with DFA (50% sampling)"
-    #SP_Summaries_welch_rect_area_5_1
-    "Total power in lowest fifth of frequencies in the Fourier power spectrum"
-    #SP_Summaries_welch_rect_centroid
-    "Centroid of the Fourier power spectrum"
-    #SB_TransitionMatrix_3ac_sumdiagcov
-    "Trace of covariance of transition matrix between symbols in 3-letter alphabet"
-    #PD_PeriodicityWang_th0_01
-    "Periodicity measure of (Wang et al. 2007)"]
+                             #DN_HistogramMode_5
+                             "Mode of z-scored distribution (5-bin histogram)"
+                             #DN_HistogramMode_10
+                             "Mode of z-scored distribution (10-bin histogram)"
+                             #CO_Embed2_Dist_tau_d_expfit_meandiff
+                             "Exponential fit to successive distances in 2-d embedding space"
+                             #CO_f1ecac
+                             "First 1/𝑒 crossing of autocorrelation function"
+                             #CO_FirstMin_ac
+                             "First minimum of autocorrelation function"
+                             #CO_HistogramAMI_even_2_5
+                             "Automutual information, 𝑚=2,𝜏=5"
+                             #CO_trev_1_num
+                             "Time-reversibility statistic, ⟨(𝑥ₜ₊₁−𝑥ₜ)³⟩ₜ"
+                             #DN_OutlierInclude_p_001_mdrmd
+                             "Time intervals between successive extreme events above the mean"
+                             #DN_OutlierInclude_n_001_mdrmd
+                             "Time intervals between successive extreme events below the mean"
+                             #FC_LocalSimple_mean1_tauresrat
+                             "Change in correlation length after iterative differencing"
+                             #FC_LocalSimple_mean3_stderr
+                             "Mean error from a rolling 3-sample mean forecasting"
+                             #IN_AutoMutualInfoStats_40_gaussian_fmmi
+                             "First minimum of the automutual information function"
+                             #MD_hrv_classic_pnn40
+                             "Proportion of successive differences exceeding 0.04𝜎 (Mietus 2002)"
+                             #SB_BinaryStats_diff_longstretch0
+                             "Longest period of successive incremental decreases"
+                             #SB_BinaryStats_mean_longstretch1
+                             "Longest period of consecutive values above the mean"
+                             #SB_MotifThree_quantile_hh
+                             "Shannon entropy of two successive letters in equiprobable 3-letter symbolization"
+                             #SC_FluctAnal_2_rsrangefit_50_1_logi_prop_r1
+                             "Proportion of slower timescale fluctuations that scale with linearly rescaled range fits"
+                             #SC_FluctAnal_2_dfa_50_1_2_logi_prop_r1
+                             "Proportion of slower timescale fluctuations that scale with DFA (50% sampling)"
+                             #SP_Summaries_welch_rect_area_5_1
+                             "Total power in lowest fifth of frequencies in the Fourier power spectrum"
+                             #SP_Summaries_welch_rect_centroid
+                             "Centroid of the Fourier power spectrum"
+                             #SB_TransitionMatrix_3ac_sumdiagcov
+                             "Trace of covariance of transition matrix between symbols in 3-letter alphabet"
+                             #PD_PeriodicityWang_th0_01
+                             "Periodicity measure of (Wang et al. 2007)"]