Seaborn — Declarative Visualization

After completing this topic

You will understand how Seaborn's declarative syntax differs from Matplotlib, and you will be able to draw box plots, count plots, and heatmaps with a single line of code.

When Matplotlib Becomes Inconvenient

To draw a "salary distribution by department" using Matplotlib, you would write code like this:

import matplotlib.pyplot as plt
import numpy as np

departments = df['department'].unique()
positions = range(len(departments))

fig, ax = plt.subplots()
for i, dept in enumerate(departments):
    salaries = df[df['department'] == dept]['salary']
    ax.boxplot(salaries, positions=[i], widths=0.6)

ax.set_xticks(positions)
ax.set_xticklabels(departments)
ax.set_ylabel('Salary')
ax.set_title('Salary Distribution by Department')
plt.show()

You have to loop, separate the data, calculate positions, and set labels directly. With Seaborn, the same graph is:

import seaborn as sns

sns.boxplot(data=df, x='department', y='salary')
plt.show()

Just two lines. You just say "what to show," and Seaborn handles the rest. This is declarative syntax.

Imperative vs. Declarative

Seaborn is a wrapper built on top of Matplotlib. It creates beautiful graphs quickly with default settings, and if you need fine-tuning, you can go down to Matplotlib.

5 Core Graphs

1. Distribution — histplot

sns.histplot(data=df, x='salary', bins=20, kde=True)

Adding kde=True overlays a density estimate line (smooth curve).

2. Distribution by Category — boxplot

sns.boxplot(data=df, x='department', y='salary')

See median, quartiles, and outliers at a glance. Most effective for comparing salaries between departments.

3. Counting — countplot

sns.countplot(data=df, x='department', hue='gender')

Shows the frequency of each category with bars. Adding a subcategory, such as gender, with hue separates it by color.

4. Relationship — scatterplot

sns.scatterplot(data=df, x='experience', y='salary', hue='department')

Plots the relationship between two numerical variables as points. At a glance, you can see whether experience and salary are proportional.

5. Correlation — heatmap

corr = df[['salary', 'experience', 'age']].corr()
sns.heatmap(corr, annot=True, cmap='coolwarm', vmin=-1, vmax=1)

Display the numbers with annot=True and specify the color palette with cmap. Positive correlation (red) and negative correlation (blue) are distinguished by color.

hue — The Third Variable

Seaborn's most powerful feature:

# Without hue — overall scatter plot
sns.scatterplot(data=df, x='experience', y='salary')

# Add hue — separate by color for each department
sns.scatterplot(data=df, x='experience', y='salary', hue='department')

hue encodes the third variable with color. To do this with Matplotlib, you would have to loop and specify the colors directly. Seaborn only needs one word.

Style Settings

# Change the overall style
sns.set_theme(style='whitegrid')

# Change the palette
sns.set_palette('Set2')

# For individual graphs
sns.boxplot(data=df, x='dept', y='salary', palette='pastel')

Style options:

• darkgrid — default, gray background + grid
• whitegrid — white background + grid
• dark — gray background without grid
• white — clean white background
• ticks — only axis ticks

Seaborn + Matplotlib Combination

fig, axes = plt.subplots(1, 2, figsize=(12, 5))

sns.boxplot(data=df, x='dept', y='salary', ax=axes[0])
axes[0].set_title('Salary by Department')

sns.histplot(data=df, x='salary', bins=20, ax=axes[1])
axes[1].set_title('Overall Salary Distribution')

plt.tight_layout()
plt.show()

You can put Seaborn graphs into Matplotlib's Axes using the ax parameter. Layout with Matplotlib, graph content with Seaborn — this combination is a practical pattern.

Key Takeaway

Seaborn is a visualization library that you simply declare "what to show," and it will draw the rest. Encoding the third variable with color with just one word, hue — this is Seaborn's core strength. Quick exploration with Seaborn, fine-tuning with Matplotlib — you need to know how to use both.