Use CairoMakie

Author: kahaaga

docs/Project.toml

@@ -1,12 +1,11 @@
 [deps]
-Conda = "8f4d0f93-b110-5947-807f-2305c1781a2d"
+CairoMakie = "13f3f980-e62b-5c42-98c6-ff1f3baf88f0"
 DelayEmbeddings = "5732040d-69e3-5649-938a-b6b4f237613f"
 Documenter = "e30172f5-a6a5-5a46-863b-614d45cd2de4"
 DocumenterTools = "35a29f4d-8980-5a13-9543-d66fff28ecb8"
 DynamicalSystems = "61744808-ddfa-5f27-97ff-6e42cc95d634"
 Distributions = "31c24e10-a181-5473-b8eb-7969acd0382f"
-PyCall = "438e738f-606a-5dbb-bf0a-cddfbfd45ab0"
-PyPlot = "d330b81b-6aea-500a-939a-2ce795aea3ee"
 StaticArrays = "90137ffa-7385-5640-81b9-e52037218182"
 StatsBase = "2913bbd2-ae8a-5f71-8c99-4fb6c76f3a91"
+Statistics = "10745b16-79ce-11e8-11f9-7d13ad32a3b2"
 Wavelets = "29a6e085-ba6d-5f35-a997-948ac2efa89a"

docs/src/estimators.md

@@ -19,8 +19,8 @@ entropy is compared with the largest Lyapunov exponents from time series of the
 logistic map. Entropy estimates using [`SymbolicWeightedPermutation`](@ref)
 and [`SymbolicAmplitudeAwarePermutation`](@ref) are added here for comparison.
 
-```@example
-using Entropies, DynamicalSystems, PyPlot
+```@example MAIN
+using DynamicalSystems, CairoMakie
 
 ds = Systems.logistic()
 rs = 3.4:0.001:4
@@ -34,48 +34,35 @@ hs_wtperm = Float64[]
 hs_ampperm = Float64[]
 
 base = Base.MathConstants.e
-
-# Original paper doesn't use random assignment for ties, here: sort after order of occurrence
-lt = Base.isless
-est = SymbolicPermutation(m = m, τ = τ, lt = lt)
-est_aa = SymbolicAmplitudeAwarePermutation(m = m, τ = τ, lt = lt)
-est_wt = SymbolicWeightedPermutation(m = m, τ = τ, lt = lt)
-
 for r in rs
     ds.p[1] = r
     push!(lyaps, lyapunov(ds, N_lyap))
 
     x = trajectory(ds, N_ent) # time series
-    hperm = Entropies.genentropy(x, est, base = base)
-    hampperm = Entropies.genentropy(x, est_aa, base = base)
+    hperm = Entropies.genentropy(x, SymbolicPermutation(m = m, τ = τ), base = base)
     hwtperm = Entropies.genentropy(x, SymbolicWeightedPermutation(m = m, τ = τ), base = base)
-    push!(hs_perm, hperm); push!(hs_ampperm, hampperm); push!(hs_wtperm, hwtperm);
+    hampperm = Entropies.genentropy(x, SymbolicAmplitudeAwarePermutation(m = m, τ = τ), base = base)
+
+    push!(hs_perm, hperm); push!(hs_wtperm, hwtperm); push!(hs_ampperm, hampperm)
 end
 
-f = figure(figsize = (6, 8))
-a1 = subplot(411)
-plot(rs, lyaps); ylim(-2, log(2)); ylabel("\$\\lambda\$")
-a1.axes.get_xaxis().set_ticklabels([])
-xlim(rs[1], rs[end]);
-
-a2 = subplot(412)
-plot(rs, hs_perm; color = "C2"); xlim(rs[1], rs[end]);
-xlabel(""); ylabel("\$h_6 (SP)\$")
-
-a3 = subplot(413)
-plot(rs, hs_wtperm; color = "C3"); xlim(rs[1], rs[end]);
-xlabel(""); ylabel("\$h_6 (SWP)\$")
-
-a4 = subplot(414)
-plot(rs, hs_ampperm; color = "C4"); xlim(rs[1], rs[end]);
-xlabel("\$r\$"); ylabel("\$h_6 (SAAP)\$")
-tight_layout()
-savefig("permentropy.png")
+fig = Figure()
+a1 = Axis(fig[1,1]; ylabel = L"\lambda")
+lines!(a1, rs, lyaps); ylims!(a1, (-2, log(2)))
+a2 = Axis(fig[2,1]; ylabel = L"h_6 (SP)")
+lines!(a2, rs, hs_perm; color = Cycled(2))
+a3 = Axis(fig[3,1]; ylabel = L"h_6 (WT)")
+lines!(a3, rs, hs_wtperm; color = Cycled(3))
+a4 = Axis(fig[4,1]; ylabel = L"h_6 (SAAP)")
+lines!(a4, rs, hs_ampperm; color = Cycled(4))
+a4.xlabel = L"r"
+
+for a in (a1,a2,a3)
+    hidexdecorations!(a, grid = false)
+end
+fig
 ```
 
-![](permentropy.png)
-
-
 ## Visitation frequency (binning)
 
 ```@docs
@@ -116,20 +103,19 @@ points are within radius `1.5` of `p`. Plotting the actual points, along with th
 associated probabilities estimated by the KDE procedure, we get the following surface 
 plot.
 
-```@example
-using Distributions, PyPlot, DelayEmbeddings, Entropies
+```@example MAIN
+using DynamicalSystems, CairoMakie, Distributions
 𝒩 = MvNormal([1, -4], 2)
 N = 500
 D = Dataset(sort([rand(𝒩) for i = 1:N]))
 x, y = columns(D)
 p = probabilities(D, NaiveKernel(1.5))
-surf(x, y, p.p)
-xlabel("x"); ylabel("y")
-savefig("kernel_surface.png")
+fig, ax = surface(x, y, p.p; axis=(type=Axis3,))
+ax.zlabel = "P"
+ax.zticklabelsvisible = false
+fig
 ```
 
-![](kernel_surface.png)
-
 ## Time-scale (wavelet)
 
 ```@docs
@@ -138,41 +124,36 @@ TimeScaleMODWT
 
 ### Example
 
-The scale-resolved wavelet entropy should be lower for very regular signals (most of the 
+The scale-resolved wavelet entropy should be lower for very regular signals (most of the
 energy is contained at one scale) and higher for very irregular signals (energy spread
 more out across scales).
 
-```@example
-using Entropies, PyPlot
+```@example MAIN
+using DynamicalSystems, CairoMakie
 N, a = 1000, 10
 t = LinRange(0, 2*a*π, N)
 
 x = sin.(t);
-y = sin.(t .+  cos.(t/0.5));
+y = sin.(t .+ cos.(t/0.5));
 z = sin.(rand(1:15, N) ./ rand(1:10, N))
 
 est = TimeScaleMODWT()
-h_x = Entropies.genentropy(x, est)
-h_y = Entropies.genentropy(y, est)
-h_z = Entropies.genentropy(z, est)
-
-f = figure(figsize = (10,6))
-ax = subplot(311)
-px = plot(t, x; color = "C1", label = "h=$(h=round(h_x, sigdigits = 5))"); 
-ylabel("x"); legend()
-ay = subplot(312)
-py = plot(t, y; color = "C2", label = "h=$(h=round(h_y, sigdigits = 5))"); 
-ylabel("y"); legend()
-az = subplot(313)
-pz = plot(t, z; color = "C3", label = "h=$(h=round(h_z, sigdigits = 5))"); 
-ylabel("z"); xlabel("Time"); legend()
-tight_layout()
-savefig("waveletentropy.png")
+h_x = genentropy(x, est)
+h_y = genentropy(y, est)
+h_z = genentropy(z, est)
+
+fig = Figure()
+ax = Axis(fig[1,1]; ylabel = "x")
+lines!(ax, t, x; color = Cycled(1), label = "h=$(h=round(h_x, sigdigits = 5))");
+ay = Axis(fig[2,1]; ylabel = "y")
+lines!(ay, t, y; color = Cycled(2), label = "h=$(h=round(h_y, sigdigits = 5))");
+az = Axis(fig[3,1]; ylabel = "z", xlabel = "time")
+lines!(az, t, z; color = Cycled(3), label = "h=$(h=round(h_z, sigdigits = 5))");
+for a in (ax, ay, az); axislegend(a); end
+for a in (ax, ay); hidexdecorations!(a; grid=false); end
+fig
 ```
 
-![](waveletentropy.png)
-
-
 ## Nearest neighbor estimators
 
 ### Kraskov
@@ -189,20 +170,20 @@ KozachenkoLeonenko
 
 #### Example
 
-This example reproduces Figure in Charzyńska & Gambin (2016)[^Charzyńska2016]. Both 
-estimators nicely converge to the true entropy with increasing time series length. 
-For a uniform 1D distribution ``U(0, 1)``, the true entropy is `0` (red line).
+This example reproduces Figure in Charzyńska & Gambin (2016)[^Charzyńska2016]. Both
+estimators nicely converge to the true entropy with increasing time series length.
+For a uniform 1D distribution ``U(0, 1)``, the true entropy is `0`.
 
-```@example
-using Entropies, DelayEmbeddings, StatsBase
-import Distributions: Uniform, Normal
+```@example MAIN
+using DynamicalSystems, CairoMakie, Statistics
+using Distributions: Uniform, Normal
 
 Ns = [100:100:500; 1000:1000:10000]
 Ekl = Vector{Vector{Float64}}(undef, 0)
 Ekr = Vector{Vector{Float64}}(undef, 0)
 
 est_nn = KozachenkoLeonenko(w = 0)
-# with k = 1, Kraskov is virtually identical to KozachenkoLeonenko, so pick a higher 
+# with k = 1, Kraskov is virtually identical to KozachenkoLeonenko, so pick a higher
 # number of neighbors
 est_knn = Kraskov(w = 0, k = 3)
 
@@ -220,25 +201,18 @@ for N in Ns
     push!(Ekr, kr)
 end
 
-# Plot
-using PyPlot, StatsBase
-f = figure(figsize = (5,6))
-ax = subplot(211)
-px = PyPlot.plot(Ns, mean.(Ekl); color = "C1", label = "KozachenkoLeonenko"); 
-PyPlot.plot(Ns, mean.(Ekl) .+ StatsBase.std.(Ekl); color = "C1", label = ""); 
-PyPlot.plot(Ns, mean.(Ekl) .- StatsBase.std.(Ekl); color = "C1", label = ""); 
-
-xlabel("Time step"); ylabel("Entropy (nats)"); legend()
-ay = subplot(212)
-py = PyPlot.plot(Ns, mean.(Ekr); color = "C2", label = "Kraskov"); 
-PyPlot.plot(Ns, mean.(Ekr) .+ StatsBase.std.(Ekr); color = "C2", label = ""); 
-PyPlot.plot(Ns, mean.(Ekr) .- StatsBase.std.(Ekr); color = "C2", label = ""); 
-
-xlabel("Time step"); ylabel("Entropy (nats)"); legend()
-tight_layout()
-PyPlot.savefig("nn_entropy_example.png")
-```
+fig = Figure()
+ax = Axis(fig[1,1]; ylabel = "entropy (nats)", title = "KozachenkoLeonenko")
+lines!(ax, Ns, mean.(Ekl); color = Cycled(1))
+band!(ax, Ns, mean.(Ekl) .+ std.(Ekl), mean.(Ekl) .- std.(Ekl);
+color = (Main.COLORS[1], 0.5))
+
+ay = Axis(fig[2,1]; xlabel = "time step", ylabel = "entropy (nats)", title = "Kraskov")
+lines!(ay, Ns, mean.(Ekr); color = Cycled(2))
+band!(ay, Ns, mean.(Ekr) .+ std.(Ekr), mean.(Ekr) .- std.(Ekr); 
+color = (Main.COLORS[2], 0.5))
 
-![](nn_entropy_example.png)
+fig
+```
 
 [^Charzyńska2016]: Charzyńska, A., & Gambin, A. (2016). Improvement of the k-NN entropy estimator with applications in systems biology. Entropy, 18(1), 13.