plots

replication/weather/weather.jl

@@ -3,15 +3,13 @@ using DataFrames
 using PyPlot
 using Random
 using Colors
-using Plots
 using Dates
 using MondrianForests
 
 # plot setup
 rcParams = PyPlot.PyDict(PyPlot.matplotlib."rcParams")
 rcParams["text.usetex"] = true
 rcParams["font.family"] = "serif"
-#rcParams["text.latex.preamble"] = "\\usepackage[sfdefault,light]{FiraSans}"
 plt.ioff()
 
 function load_data(; limit=nothing)
@@ -172,9 +170,9 @@ function plot_debiased_forest(trees1, trees2, ax)
     Y = [data[i, :RainTomorrow] for i in 1:nrow(data)]
     n = length(X)
     all_counts1 = [[sum(MondrianForests.is_in(X[i], cell) for i in 1:n)
-                   for cell in all_cells1[j]] for j in 1:length(trees1)]
+                    for cell in all_cells1[j]] for j in 1:length(trees1)]
     all_counts2 = [[sum(MondrianForests.is_in(X[i], cell) for i in 1:n)
-                   for cell in all_cells2[j]] for j in 1:length(trees2)]
+                    for cell in all_cells2[j]] for j in 1:length(trees2)]
     all_ones1 = [[sum(MondrianForests.is_in(X[i], cell) * Y[i] for i in 1:n)
                   for cell in all_cells1[j]] for j in 1:length(trees1)]
     all_ones2 = [[sum(MondrianForests.is_in(X[i], cell) * Y[i] for i in 1:n)
@@ -188,18 +186,22 @@ function plot_debiased_forest(trees1, trees2, ax)
     for c in 1:length(refined_cells)
         r_cell = refined_cells[c]
         x = MondrianForests.get_center(r_cell)
-        cell_ids1 = [[c for c in 1:length(all_cells1[j]) if
-                     MondrianForests.is_in(x, all_cells1[j][c])][] for j in 1:length(trees1)]
-        cell_ids2 = [[c for c in 1:length(all_cells2[j]) if
-                     MondrianForests.is_in(x, all_cells2[j][c])][] for j in 1:length(trees2)]
+        cell_ids1 = [[c
+                      for c in 1:length(all_cells1[j])
+                      if
+                      MondrianForests.is_in(x, all_cells1[j][c])][] for j in 1:length(trees1)]
+        cell_ids2 = [[c
+                      for c in 1:length(all_cells2[j])
+                      if
+                      MondrianForests.is_in(x, all_cells2[j][c])][] for j in 1:length(trees2)]
         counts1 = [all_counts1[j][cell_ids1[j]] for j in 1:length(trees1)]
         counts2 = [all_counts2[j][cell_ids2[j]] for j in 1:length(trees2)]
         ones1 = [all_ones1[j][cell_ids1[j]] for j in 1:length(trees1)]
         ones2 = [all_ones2[j][cell_ids2[j]] for j in 1:length(trees2)]
         rs1 = ones1 ./ counts1
         rs2 = ones2 ./ counts2
         if !all(isnan.(rs1))
-            # TODO is this right
+            # TODO is this right, fix in other non-debiased version too
             ratios1[c] = sum(r for r in rs1 if !isnan(r)) / sum(1 for r in rs1 if !isnan(r))
         else
             ratios1[c] = NaN
@@ -293,7 +295,7 @@ function make_forest_design_plot(data, trees, x_min, x_max,
 end
 
 function make_debiased_forest_design_plot(data, trees1, trees2, x_min, x_max,
-                                 y_min, y_max, design_points, filename)
+                                          y_min, y_max, design_points, filename)
     # TODO
     (fig, ax) = plt.subplots(figsize=figsize)
     plot_debiased_forest(trees1, trees2, ax)
@@ -309,7 +311,6 @@ function make_debiased_forest_design_plot(data, trees1, trees2, x_min, x_max,
 end
 
 # get data and plot params
-#limit = 1000
 limit = nothing
 (data, x_min, x_max, y_min, y_max) = load_data(limit=limit)
 dry_color = "#da6200"
@@ -339,7 +340,6 @@ for i in 1:length(seeds)
     push!(trees, tree)
 end
 
-#=
 # plot data
 println("plotting data")
 filename = "./replication/weather/weather_data.png"
@@ -362,7 +362,6 @@ for i in [2]
     global filename = "./replication/weather/weather_forest_" * string(i) * ".png"
     make_forest_plot(data, trees[1:i], x_min, x_max, y_min, y_max, filename)
 end
-=#
 
 # plot debiased forest with design points
 i = 15
@@ -382,11 +381,9 @@ global filename = "./replication/weather/weather_debiased_forest_design.png"
 make_debiased_forest_design_plot(data, trees[1:i], trees2[1:i], x_min, x_max,
                                  y_min, y_max, design_points, filename)
 
-#=
 i = 30
 # plot forest with design points
 println("plotting forest with ", i, " trees and design points")
 global filename = "./replication/weather/weather_forest_design.png"
 make_forest_design_plot(data, trees[1:i], x_min, x_max, y_min,
                         y_max, design_points, filename)
-=#

replication/weather/weather_cv.jl

@@ -11,7 +11,6 @@ using MondrianForests
 rcParams = PyPlot.PyDict(PyPlot.matplotlib."rcParams")
 rcParams["text.usetex"] = true
 rcParams["font.family"] = "serif"
-#rcParams["text.latex.preamble"] = "\\usepackage[sfdefault,light]{FiraSans}"
 plt.ioff()
 
 function load_data(; limit=nothing)

shell.nix

@@ -11,6 +11,6 @@ in pkgs.mkShell {
     ];
     shellHook = ''
         # run this to link Julia PyCall package to nixpkgs python3
-        #julia --color=yes -e 'using Pkg; ENV["PYTHON"]="${pkgs.python3}/bin/python3"; Pkg.build("PyCall")'
+        julia --color=yes -e 'using Pkg; ENV["PYTHON"]="${pkgs.python3}/bin/python3"; Pkg.build("PyCall")'
     '';
 }