X4DS

This repo is a collection of

Python & Julia port of codes in the following excellent R books:

• An Introduction to Statistical Learning (ISLR)
• Statistical Rethinking (SRT)

Code Styles

2.1. Basics

• prefer enumerate() over range(len())

xs = range(3)

# good
for ind, x in enumerate(xs):
  print(f'{ind}: {x}')

# bad
for i in range(len(xs)):
  print(f'{i}: {xs[i]}')

2.2. Matplotlib

including seaborn

• prefer Axes object over Figure object
• use constrained_layout=True when draw subplots

# good
_, axes = plt.subplots(1, 2, constrained_layout=True)
axes[0].plot(x1, y1)
axes[1].hist(x2, y2)

# bad
plt.subplot(121)
plt.plot(x1, y1)
plt.subplot(122)
plt.hist(x2, y2)

• prefer axes.flatten() over plt.subplot() in cases where subplots' data is iterable
• prefer zip() or enumerate() over range() for iterable objects

# good
_, ax = plt.subplots(2, 2, figsize=[12,8],constrained_layout=True)

for ax, x, y in zip(axes.flatten(), xs, ys):
  ax.plot(x, y)

# bad
for i in range(4):
  ax = plt.subplot(2, 2, i+1)
  ax.plot(x[i], y[i])

• prefer set() method over set_*() method

# good
ax.set(xlabel='x', ylabel='y')

# bad
ax.set_xlabel('x')
ax.set_ylabel('y')

• Prefer despine() over ax.spines[*].set_visible()

# good
sns.despine()

# bad
ax.spines["top"].set_visible(False)
ax.spines["bottom"].set_visible(False)
ax.spines["right"].set_visible(False)
ax.spines["left"].set_visible(False)

2.3. Pandas

• prefer df['col'] over df.col

# good
movies['duration']

# bad
movies.duration

• prefer df.query over df[] or df.loc[] in simple-selection

# good
movies.query('duration >= 200')

# bad
movies[movies['duration'] >= 200]
movies.loc[movies['duration'] >= 200, :]

• prefer df.loc and df.iloc over df[] in multiple-selection

# good
movies.loc[movies['duration'] >= 200, 'genre']
movies.iloc[0:2, :]

# bad
movies[movies['duration'] >= 200].genre
movies[0:2]

LaTeX Styles

Multiple lines

Reduce the use of begin{array}...end{array}

• equations: begin{aligned}...end{aligned}

$$
\begin{aligned}
y_1 = x^2 + 2*x \\
y_2 = x^3 + x
\end{aligned}
$$

• equations with conditions: begin{cases}...end{cases}

$$
\begin{cases}
y = x^2 + 2*x & x > 0 \\
y = x^3 + x & x ≤ 0
\end{cases}
$$

• matrix: begin{matrix}...end{matrix}

$$
\begin{vmatrix}
  a + a^′ & b + b^′ \\ c & d
  \end{vmatrix}= \begin{vmatrix}
  a & b \\ c & d
  \end{vmatrix} + \begin{vmatrix}
  a^′ & b^′ \\ c & d
\end{vmatrix}
$$

Brackets

• prefer \Bigg...\Bigg over \left...\right

$$
A\Bigg[v_1\ v_2\ ⋯\ v_r\Bigg]
$$

• prefer \underset{}{} over \underset{}

$$
\underset{θ}{\mathrm{argmax}}\ p(x_i|θ)
$$

Expressions

• prefer ^{\top} over ^T for transpose

$$ 𝐀^⊤ $$

$$
𝐀^{\top}
$$

• prefer \to over \rightarrow for limit

$$ \lim_{n → ∞} $$

$$
\lim_{n\to \infty}
$$

• prefer underset{}{} over \limits_

$$ \underset{w}{\rm argmin}\ (wx +b) $$

$$
\underset{w}{\rm argmin}\ (wx +b)
$$

Fonts

• prefer \mathrm over \mathop or \operatorname

$$
θ_{\mathrm{MLE}}=\underset{θ}{\mathrm{argmax}}\ ∑_{i = 1}^{N}\log p(x_i|θ)
$$

ISLR

• Chapter 02 - Statistical Learning
• Chapter 03 - Linear Regression
• Chapter 04 - Classification
• Chapter 05 - Resampling Methods
• Chapter 06 - Linear Model Selection and Regularization
• Chapter 07 - Moving Beyond Linearity
• Chapter 08 - Tree Based Methods
• Chapter 09 - Support Vector Machines
• Chapter 10 - Unsupervised Learning

References

• ISL-python
• ISLR.jl
• handson-ml2