`plotting`

Utility Functions that can be used to manage colors and other plotting tools. Includes all stable utility functions.

Work with color maps

matplotlib.colormaps['tab10']

tab10

under

bad

over

print(cmaps['Cyclic'])
matplotlib.colormaps[cmaps['Cyclic'][2]]

['twilight', 'twilight_shifted', 'hsv']

hsv

under

bad

over

print(cmaps['Sequential'])
matplotlib.colormaps[cmaps['Sequential'][-1]]

['Greys', 'Purples', 'Blues', 'Greens', 'Oranges', 'Reds', 'YlOrBr', 'YlOrRd', 'OrRd', 'PuRd', 'RdPu', 'BuPu', 'GnBu', 'PuBu', 'YlGnBu', 'PuBuGn', 'BuGn', 'YlGn']

YlGn

under

bad

over

source

plot_cmap_collections

def plot_cmap_collections(
    cmap_collections:str | list[str]=None, # list of color map collections to display (from cmaps.keys())
):

Plot all color maps in the collections passed as cmap_collections

The following color map collections are defined:

- Cyclic
- Diverging
- Miscellaneous
- Perceptually Uniform Sequential
- Qualitative
- Sequential
- Sequential (2)

plot_cmap_collections will plot a color bar for each color map in the selected collections:

A single collection

plot_cmap_collections('Cyclic')

Several collections

plot_cmap_collections(['Qualitative', 'Sequential'])

All the collections

plot_cmap_collections()

source

plot_color_bar

def plot_color_bar(
    cmap:str, # string name of one of the cmaps
    series:list[Union[int, float]]=None, # series of numerical values to show for each color
):

Plot a color bar with value overlay from series based on cmap

Technical note: Adapts the label color to the background color for readibility. To do so it computes the relative luminance as: L = 0.2126 * R + 0.7152 * G + 0.0722 * B Use black text when L > 0.5, otherwise white text.

plot_color_bar('tab10', range(10))

plot_color_bar('tab10', series=range(6))

plot_color_bar('tab10', series=[0, 1, 2])

plot_color_bar('Greens', range(17))

source

get_color_mapper

def get_color_mapper(
    series:list[Union[int, float]], # series of values to map to colors
    cmap:str='tab10', # name of the cmap to use
):

Return color mapper based on a color map and a series of values

Usage

This function is used to ensure coherent colors for different plots.

Define a color mapper based on values and cmap: clr_mapper = get_color_mapper([1, 2, 3, 4], cmap='Paired)
Call the color mapper and have it return the appropriate values for any plot: clr_mapper.to_rgba(2)

Example

Let’s take a dataset \((X,y)\), generated by sklearn using make_blob for 5,000 samples with 6 features each and grouped in 10 centers. The \(X\) and \(y\) are saved as npy file

import pandas as pd
with open('data-dev/ten-blobs-6-cols-X.npy', 'rb') as fp:
    X = np.load(fp)
with open('data-dev/ten-blobs-6-cols-y.npy', 'rb') as fp:
    y = np.load(fp)
    
n_feats = X.shape[1]
col_list = [f"col_{i}" for i in range(n_feats)]

X = pd.DataFrame(X, columns=col_list)
X.head(3)

	col_0	col_1	col_2	col_3	col_4	col_5
0	8.714921	0.580067	6.817529	-4.891837	-0.418419	-4.127050
1	-0.322263	-3.901781	-8.588875	1.114151	4.908420	-2.475176
2	-7.798702	-4.276232	-1.145314	-6.619232	-9.232606	6.994045

1. Define a color mapper based on values and cmap

We have clustered the 5,000 sample of \(X\) into 10 clusters (saved into a clusters.npy file)

Now we want to make a scatter plot of two of the features, displaying the 10 cluster using a cmap.

To ensure that we can keep the same cluster color mapping for other plots, we use clr_mapper to predefine how colors are mapped to each cluster:

clr_mapper = get_color_mapper(cluster_ids, cmap=cmap).

# Load clustered data from the blobs
with open('data-dev/ten-blobs-6-cols-clusters.npy', 'rb') as fp:
    clusters = np.load(fp)
    
cluster_ids = np.unique(clusters)
cluster_ids

array([0, 1, 2, 3, 4, 5, 6, 7, 8, 9], dtype=int32)

cmap='tab10'

fig, (ax1, ax2) = plt.subplots(nrows=1, ncols=2, figsize=(14, 3))

im1 = ax1.scatter(X.col_0, X.col_1, c=clusters, s=2, cmap=cmap)
ax1.set_title('2 first features, colored by cluster value')
fig.colorbar(im1)

im2 = ax2.scatter(X.col_4, X.col_5, c=clusters, s=2, cmap=cmap)
ax2.set_title('2 last features, colored by cluster value')
fig.colorbar(im2)
plt.show()

clr_mapper = get_color_mapper(cluster_ids, cmap=cmap)

Call the color mapper and use it in any plot

Use for another plot, showing another feature, and its value for each sample, colored according to its cluster

featname = 'col_4'
plt.figure(figsize=(12, 3))
plt.plot(X[featname], c='grey', alpha=.66, lw=0.25)
plt.title(f'{featname}.')
plt.show()

plt.figure(figsize=(12, 3))
plt.plot(X[featname], c='grey', alpha=.5, lw=0.1)
for c in cluster_ids:
    mask = y == c
    X[f"{featname}_{c}"] = X.loc[:, featname]
    X.loc[~mask, f"{featname}_{c}"] = np.nan
    plt.plot(X[f"{featname}_{c}"], label=str(c), c=clr_mapper.to_rgba(c), lw=0, marker='o', markersize=1)
plt.title(f'{featname}. Data points colored according to the cluster it belongs to.')
plt.legend()
plt.show()

Advanced plots

source

plot_feature_scatter

def plot_feature_scatter(
    X:np.ndarray, # input dataset. `X.shape[1]` is used to set the total number of features
    y:Optional[np.ndarray]=None, # target dataset
    n_plots:int=2, # number of feature pairs scatter plot to show
    axes_per_row:int=3, # number of axes per row. number of rows will be calculated accordingly
    axes_size:int=5, # size of one axes. figsize will be (ncols * axes_size, nrows * axes_size)
):

Plots n_plots scatter plots of randomly selected combinations of two features out of X

X.shape

(5000, 16)

with open('data-dev/ten-blobs-6-cols-X.npy', 'rb') as fp:
    X = np.load(fp)
with open('data-dev/ten-blobs-6-cols-y.npy', 'rb') as fp:
    y = np.load(fp)

n_feats = X.shape[1]
col_list = [f"col_{i}" for i in range(n_feats)]
X = pd.DataFrame(X, columns=col_list)

plot_feature_scatter(X.values, y, n_plots=6, axes_per_row=3, axes_size=5)

When not value is available for y, it is set to 1 by default

plot_feature_scatter(X.values, n_plots=4, axes_per_row=2, axes_size=2)