Seaborn Heatmap with logarithmic-scale colorbar

Question

Is there a way to set the color bar scale to log on a seaborn heat map graph?

I am using a pivot table output from pandas as an input to the call

sns.heatmap(df_pivot_mirror, annot=False, xticklabels=256, yticklabels=128, cmap=plt.cm.YlOrRd_r)

cphlewis · Answer 1 · 2021-06-15T00:31:51.600

53

If you have a current install of seaborn, norm=LogNorm() in the call to heatmap works now. (Pointed out in the comments -- thank you.) Adding this to one of the seaborn examples:

import numpy as np
import seaborn as sns; sns.set_theme(style='white')
import matplotlib.pyplot as plt
from matplotlib.colors import LogNorm, Normalize
from matplotlib.ticker import MaxNLocator

flights = sns.load_dataset("flights")
flights = flights.pivot("month", "year", "passengers")


f3, ax5 = plt.subplots(1,1)
sns.heatmap(flights, square=True, norm=LogNorm())

You can pass through colorbar arguments as keywords in the seaborn wrapper, but they sometimes collide with the seaborn choices:

sns.heatmap(flights, square=True, norm=LogNorm(), cbar_kws={'ticks':MaxNLocator(2), 'format':'%.e'})

For comparison, this is the matplotlib heatmap without seaborn's improvements -- the colorbar arguments have both been applied:

f5, ax6 = plt.subplots(1,1)
im6 = plt.imshow(flights, norm=LogNorm())
cbar6 = ax.figure.colorbar(im6, ax=ax6, ticks=MaxNLocator(2), format='%.e')

If you have to use an older install and LogNorm doesn't work in seaborn, see the previous versions of this answer for a workaround.

edited Jun 15 '21 at 00:31

answered Apr 29 '16 at 00:20

cphlewis

15,759
4
46
55

3

seaborn issue https://github.com/mwaskom/seaborn/issues/459 also makes it fixable (a different way) – cphlewis May 06 '16 at 08:31
1

The only thing is if some of your values are negative, then that's a problem. Hyperbolic inverse sin (arcsinh) works better for negative and 0 values. – wordsforthewise Jan 04 '18 at 04:57
2

There isn't an arcsinh built-in to matplotlib, but you can use SymLogNorm for negative values. Not sure if it handles 0's though. https://matplotlib.org/devdocs/tutorials/colors/colormapnorms.html#sphx-glr-tutorials-colors-colormapnorms-py – wordsforthewise Jan 04 '18 at 05:01
Is there a way to enable the logarithmic-color scale in a 2D histogram done by `sns.jointplot`? – Stefano Oct 05 '20 at 19:53
A log scale in [a distplot like this?](https://seaborn.pydata.org/generated/seaborn.jointplot.html) I think that deserves to be another question. – cphlewis Oct 09 '20 at 01:59
3

I just wanted to add that the issue with the hard-coded tick-locator seems to have been fixed in recent versions of seaborn. The only thing I had to do was to specify `norm=LogNorm()` and that resulted in logarithmic ticks as well. – Raven Dec 02 '20 at 09:40
You can use `LogNorm(clip=True)` to map 0s to the bottom of the colorscale. – thomaskeefe Oct 12 '22 at 18:21
1

in case anyone else ends up here looking for sns.histplot, you have to pass `vmin =None` and `vmax = None`, otherwise you get an error. – Jazz Weisman Feb 28 '23 at 00:16

Tomas G. · Answer 2 · 2020-10-14T12:19:16.213

27

Short Answer:

from matplotlib.colors import LogNorm

sns.heatmap(df, norm=LogNorm())

edited Oct 14 '20 at 12:19

answered Mar 30 '20 at 13:43

Tomas G.

3,784
25
28

score 16 · Answer 3 · answered May 19 '18 at 20:58

You can normalize the values on the colorbar with matplotlib.colors.LogNorm. I also had to manually set the labels in seaborn and ended up with the following code:

#!/usr/bin/env python3

import math

import numpy as np
import seaborn as sn
from matplotlib.colors import LogNorm

data = np.random.rand(20, 20)

log_norm = LogNorm(vmin=data.min().min(), vmax=data.max().max())
cbar_ticks = [math.pow(10, i) for i in range(math.floor(math.log10(data.min().min())), 1+math.ceil(math.log10(data.max().max())))]

sn.heatmap(
    data,
    norm=log_norm,
    cbar_kws={"ticks": cbar_ticks}
)

heatmap rand

score 5 · Answer 4 · edited Jan 10 '20 at 12:55

Responding to cphlewis (I don't have enough reputation), I solved this problem using cbar_kws; as I saw here: seaborn clustermap: set colorbar ticks.

For example cbar_kws={"ticks":[0,1,10,1e2,1e3,1e4,1e5]}.

from matplotlib.colors import LogNorm
s=np.random.rand(20,20)
sns.heatmap(s, norm=LogNorm(s.min(),s.max()),
            cbar_kws={"ticks":[0,1,10,1e2,1e3,1e4,1e5]},
            vmin = 0.001, vmax=10000)
plt.show()

Have a nice day.

Seaborn Heatmap with logarithmic-scale colorbar

4 Answers4

Linked

Related