PlotRecipes

Primary author: Thomas Breloff (@tbreloff)

This repo maintains a collection of recipes for machine learning, graph analysis, finance, and more. It uses the powerful machinery of Plots and RecipesBase to turn simple transformations into flexible visualizations.

PlotRecipes also exports the recipes in StatPlots.jl, which is a collection of statistics recipes, including functionality for DataFrames and Distributions.

Examples

Graphs

Spectral

using PlotRecipes
n = 15
A = Float64[(rand()<0.5 ? 0 : rand()) for i=1:n,j=1:n]
for i=1:n
    A[i,1:i-1] = A[1:i-1,i]
end
node_weights = 1:n

graphplot(A,
    node_weights = 1:n,
    marker = (:heat, :rect),
    line = (3, 0.5, :blues),
    marker_z = 1:n,
    names = 1:n
)

graphplot(A,
    node_weights = 1:n,
    dim = 3,
    line = (3, 0.5, :blues),
    marker_z = 1:n
)

Arc and chord diagrams

using PlotRecipes

adjmat = Symmetric(sparse(rand(0:1,8,8)));

plot(
    graphplot(adjmat, method=:chorddiagram),
    graphplot(adjmat, method=:arcdiagram, markersize=3)
    )

Fun with algos. Visualizing a stress-driven layout algorithm

Julia code -- AST

using PlotRecipes
pyplot(ma=0.8,lc=:white,mc=:white,size=(1000,800))
theme(:dark)

code = :(
function transform!{T}(act::Activation{:softmax,T})
    val = forward!(act.input)
    out = act.output.val
    for i=1:act.n
        out[i] = exp(val[i])
    end
    s = one(T) / sum(out)
    for i=1:act.n
        out[i] *= s
    end
    out
end
)

plot(code, fontsize=11, shorten=0.2, axis_buffer=0.05)

Julia Type Trees

using PlotRecipes, Learn
pyplot(size=(800,500))
theme(:dark)
plot(Learnable, method=:tree)

Maps and Shapefile.jl

using PlotRecipes, Shapefile
dir = "https://github.com/nvkelso/natural-earth-vector/raw/master/110m_physical/"
fn = "ne_110m_land.shp"
run(`wget $dir/$fn -P /tmp/`)
shp = open("/tmp/$fn") do fd
    read(fd, Shapefile.Handle)
end
plot(shp)

Portfolio Composition Maps

using PlotRecipes
tickers = ["IBM", "Google", "Apple", "Intel"]
N = 10
D = length(tickers)
weights = rand(N,D)
weights ./= sum(weights, 2)
returns = sort!((1:N) + D*randn(N))

portfoliocomposition(weights, returns, labels = tickers')

StatPlots.jl

corrplot

using PlotRecipes
M = randn(1000,4)
M[:,2] += 0.8sqrt(abs(M[:,1])) - 0.5M[:,3] + 5
M[:,3] -= 0.7M[:,1].^2 + 2

corrplot(M, label = ["x$i" for i=1:4])

Marginal Histograms

using PlotRecipes, Distributions
n = 1000
x = rand(Gamma(2), n)
y = -0.5x + randn(n)

marginalhist(x, y, fc=:plasma, bins=40)

AndrewsPlot

AndrewsPlots are a way to visualize structure in high-dimensional data by depicting each row of an array or table as a line that varies with the values in columns. https://en.wikipedia.org/wiki/Andrews_plot

using RDatasets, PlotRecipes
iris = dataset("datasets", "iris")
@df iris andrewsplot(:Species, cols(1:4), legend = :topleft)

sschelm / PlotRecipes.jl