Plotting Ungrouped Scalars¶

Here there are the classes that plot ungrouped scalars

class plotszoo.scalars.ScalarsParallelCoordinates(data, groups, target)¶

Plot a parallel coordinates plot with respect to groups and using target as rightmost field

Args:

data: plotszoo.data.DataCollection with some scalars
groups: columns of data scalars to plot the data against
target: column of data, plotted rightmost and used to color the plot

Example:

import numpy as np
import pandas as pd
import os
import matplotlib.pyplot as plt
from sklearn.datasets import load_iris
import plotszoo


iris = load_iris()
scalars = pd.DataFrame(data= np.c_[iris["data"], iris["target"]], columns=iris["feature_names"] + ["target"])

data = plotszoo.data.DataCollection()
data.set_scalars(scalars)

fig, axes = plt.subplots(1, len(iris["feature_names"]), sharey=False)

parallel_plot = plotszoo.scalars.ScalarsParallelCoordinates(data, iris["feature_names"], "target")

parallel_plot.plot(axes, cmap="tab10")

fig.set_size_inches(20, 10)
fig.savefig(os.path.join(os.path.dirname(os.path.realpath(__file__)), "images/ScalarsParallelCoordinates.png"))

plot(axes, ticks=6, adjust_whitespaces=True, cmap='Blues', cmap_fn=None, xticks_fn=None, colorbar=False)¶

Plot the parallel coordinates chart

Args:

axes: List of matplotlib axes to plot to (you must use the same number of axes and groups)
ticks: Ticks configuration dictionary or number of ticks to show in the axes (Default: 6)
cmap: matplotlib colormap to use (Default: Blues)
colorbar: Plot the colorbar (Default: False)
cmap_fn: Function to use instead of the matplotlib colormap (Default: None)
xticks_fn: Function to create the xticks (Default: None)
adjust_withspaces: Call plt.subplots_adjust(wspace=0) to make the plot prettier (can have side-effects) (Default: True)

Configuration Dictionary:

A dictionary with a key for each group. Each element is a dictionary with:

type

categorical or numeral

ticks

Number of ticks if type is numeral, the list of ticks in the type is categorical

scale

The scale to use linear or logarithmic if type is numeral, relative or sequential if the type is categorical

Configuration Dictionary Example:

{
    "config/train_steps": dict(type="categorical", ticks=[8, 16, 32, 64], scale="relative"),
    "config/gamma": dict(type="categorical", ticks=[0.9, 0.99, 0.999], scale="sequential"),
    "config/lr": dict(type="numeral", ticks=6, scale="logarithmic"),
    "config/max_clip_norm": dict(type="numeral", ticks=6, scale="linear"),
    "config/ent_coef": dict(type="numeral", ticks=6, scale="logarithmic"),
    "config/vf_coef": dict(type="numeral", ticks=6, scale="linear"),
    "config/target_entropy": dict(type="numeral", ticks=6, scale="linear"),
    "summary/reward": dict(type="numeral", ticks=6, scale="linear")
}

class plotszoo.scalars.ScalarsScatterCumulative(data, x, y, cumulative_fn=<function amax>)¶

Plot a scatterplot with some running cumulative function (ex. maximum, minimum, mean, sum, …)

Args:

data: plotszoo.data.DataCollection with some scalars
x: Index to use as x in the scatter plot (set to None to use the DataFrame index)
y: Index to use as y in the scatter plot (required)
cumulative_fn: Cumulative function to use (Default: np.max)

Example:

import numpy as np
import pandas as pd
from scipy.stats import norm
import os
import matplotlib.pyplot as plt
import plotszoo

x = np.linspace(norm.ppf(0.01), norm.ppf(0.99), 20)
pdf = pd.DataFrame(norm.pdf(x)/norm.pdf(x).sum(), columns=["pdf"])

data = plotszoo.data.DataCollection()
data.set_scalars(pdf)

fig, ax = plt.subplots()

scatter_cumulative = plotszoo.scalars.ScalarsScatterCumulative(data, x=None, y="pdf", cumulative_fn=np.sum)

scatter_cumulative.plot(ax)

fig.savefig(os.path.join(os.path.dirname(os.path.realpath(__file__)), "images/ScalarsScatterCumulative.png"))

plot(ax, sort=False, scatter=True, label=None, x_seq=False)¶

Plot the cumulative scatter plot

Args:

ax: matplotlib ax to plot to
sort: Sort the x data (Default: False)
scatter: Scatter-plot the data points (Default: True)
label: Label to use to plot (Default: None)
x_seq: Use an integer sequence as x instead of the actual values (Default: False)