2D plots — line, scatter, bar, histogram, contour, imshow
WHY these six exist (the decision tree)
WHAT distinguishes the six: what they assume about your data.
| Plot | Data shape | Implied meaning | Key question |
|---|---|---|---|
plot |
y vs ordered x | points are connected (a function/series) | "trend?" |
scatter |
(x, y) pairs | points are independent | "correlation/cluster?" |
bar |
category → value | discrete groups | "which is biggest?" |
hist |
one array of samples | frequency of value ranges | "what's the distribution?" |
contour |
grid Z[i,j]=f(x,y) | a height map (level curves) | "shape of surface?" |
imshow |
2D array of values | colored pixel grid | "the matrix/image itself?" |
1. Line plot — ax.plot
import numpy as np
import matplotlib.pyplot as plt
x = np.linspace(0, 2*np.pi, 200) # 200 pts -> smooth
y = np.sin(x)
fig, ax = plt.subplots()
ax.plot(x, y, color="tab:blue", lw=2, label="sin x")
ax.set_xlabel("x (rad)"); ax.set_ylabel("sin x")
ax.legend(); ax.grid(True)
plt.show()2. Scatter — ax.scatter
rng = np.random.default_rng(0)
x = rng.normal(size=100)
y = 2*x + rng.normal(size=100) # correlated
ax.scatter(x, y, s=30, c=x, cmap="viridis", alpha=0.7)3. Bar — ax.bar
labels = ["A", "B", "C"]; vals = [5, 9, 3]
ax.bar(labels, vals, width=0.6, color="tab:orange")WHY categorical x: Matplotlib places category at integer position and prints the label under it.
4. Histogram — ax.hist
data = rng.normal(loc=0, scale=1, size=5000)
ax.hist(data, bins=40, density=True, color="tab:green", alpha=0.8)5. Contour — ax.contour / ax.contourf
xs = np.linspace(-3, 3, 100); ys = np.linspace(-3, 3, 100)
X, Y = np.meshgrid(xs, ys)
Z = np.exp(-(X**2 + Y**2)) # a Gaussian bump
cs = ax.contour(X, Y, Z, levels=8, cmap="plasma")
ax.clabel(cs, inline=True) # label each level
# ax.contourf(...) fills between levels with color instead of lines6. imshow — ax.imshow
ax.imshow(Z, origin="lower", extent=[-3,3,-3,3], cmap="inferno", aspect="auto")
fig.colorbar(ax.images[0], ax=ax, label="f(x,y)")Putting it together (the cheat figure)
Recall Feynman: explain to a 12-year-old
Think of graph paper. Line: I plot a few dots and connect them like a join-the-dots puzzle — good for "how did temperature change all day." Scatter: I just sprinkle dots, no lines — good for "do taller kids weigh more?" Bar: tall blocks for each thing — "who scored most goals?" Histogram: I sort everyone's height into boxes (140-150cm, 150-160cm...) and count how many in each box — shows the shape of the group. Contour: like a treasure map showing hills with rings. imshow: I color every little square of graph paper by its number, like a paint-by-numbers — that's how a photo or a heatmap works.
Flashcards
When do you use scatter instead of line?
A line plot looks smooth — what actually makes it smooth?
Difference between a bar chart and a histogram?
Histogram bin width formula for B equal bins?
What does density=True do in hist and why?
What is a contour line at level c, mathematically?
Why must you call np.meshgrid before contour?
What does imshow show and how are rows/cols mapped?
Why might an imshow image appear upside-down and the fix?
origin="upper" puts row 0 at top; use origin="lower" for math orientation.Why prefer ax.plot(...) over plt.plot(...)?
What does extent do in imshow?
5-step ritual to make any plot?
Connections
- Matplotlib Figure vs Axes object model
- NumPy meshgrid and broadcasting
- Probability density functions and normalization
- Colormaps and perceptual uniformity (viridis)
- 3D surface plots (plot_surface)
- Seaborn — statistical plotting wrappers
- Image arrays and pixel coordinates
Concept Map
Hinglish (regional understanding)
Intuition Hinglish mein samjho
Dekho, har 2D plot basically ek hi cheez karta hai — numbers ko picture me badalna. Bas farq yeh hai ki tum data ko kaise samajhna chahte ho. Agar koi cheez time ke saath badal rahi hai (ordered x), to line plot — dots ko jodke trend dikhata hai. Agar points independent hain, jaise "lambe log zyada heavy hote hain kya?", to scatter use karo, koi line mat jodo. Bar chart categories ke liye hai — "kis ne sabse zyada goal maare". Aur histogram sabse zyada confuse karta hai: yeh bar chart NAHI hai tumhare data ka, yeh batata hai ki kitne samples kis range (bin) me aaye — yaani data ki shape/distribution.
Do advanced wale: contour ek mountain ki tarah hai — ko alag-alag heights pe slice karke rings banata hai, bilkul map me elevation lines jaise. Iske liye pehle np.meshgrid se 2D grid banana zaroori hai, warna error aayega. Aur imshow matrix ya image ko colored squares me dikhata hai — har cell ka value ek color ban jaata hai (heatmap). Yaad rakho: imshow me default origin upar hota hai, isliye math matrix dikhane ke liye origin="lower" lagao, warna image ulti dikhegi.
Practical advice: hamesha fig, ax = plt.subplots() use karo aur ax.plot(...) likho, plt.plot nahi — kyunki multiple subplots me explicit control chahiye. Aur axes ko label karna mat bhoolna (units ke saath), warna plot useless hai. 80/20 rule: real life me 90% kaam line + scatter + histogram se ho jaata hai, baaki teen jab specifically surface ya image dikhana ho tabhi.