galenseilis

def is_contiguous_dates(dates):
    start_date = dates.min()
    end_date = dates.max()
    date_range = pd.date_range(start=start_date, end=end_date)
    diff = date_range.difference(dates)
    if diff.size:
        return False
    else:
        return True

galenseilis

import pandas as pd

def fill_missing_dates(df, date_column='ds', value_column='y', group_column=None):
    """
    Fill missing dates between the min and max date within each group
    and assign the value of 0 for the specified column.

    Parameters:
    - df (pd.DataFrame): Input DataFrame.
    - date_column (str): Name of the datetime column.

galenseilis

import matplotlib.pyplot as plt
import numpy as np

m = 300

u1 = np.random.uniform(0.5, 4.5, size=m)
u2 = 2 / u1 + np.random.normal(scale=0.09, size=m)


plt.scatter(u1, u2)

galenseilis

import matplotlib.pyplot as plt
import statsmodels.api as sm
import statsmodels.formula.api as smf

cars = sm.datasets.get_rdataset('cars')
cars = cars.raw_data

model = smf.ols('dist ~ speed', data=cars)
results = model.fit()

galenseilis

# Tensorflow
import tensorflow as tf

print(tf.math.log(0.01 ** 200))
print(200 * tf.math.log(0.01))

# PyMC
import numpy as np
print(np.log(0.01 ** 200))
print(200 * np.log(0.01))

galenseilis

# Tensorflow
import tensorflow as tf

x = tf.range(1, 3, dtype='float32')
x *= 10
x = tf.math.log(x)
x = tf.reduce_sum(x)
x = tf.math.exp(x)

# PyMC

galenseilis

On some systems I find that python is not a defined command, but python3 is. Since Python 2 is no longer supported, and will hopefully become scarcer in dependencies, we can assume that python represents the latest installed Python.

Assuming python3 is defined, we can define our alias in a permanent way by:

1. Open config: vim ~/.bashrc
2. Add the line alias python='python3'
3. Save and exit (:wq)
4. Run . ~/.bashrc from home.

galenseilis

print("All models are wrong, but some are useful.")

galenseilis

import matplotlib.pyplot as plt
import numpy as np

L = np.random.multinomial(1, pvals=[1/3]*3, size=1000)
Q = L @ np.diag([1, 7, 31])
M = np.sum(Q, axis=1)

X = np.random.normal(loc=M, size=M.size)

galenseilis

from itertools import product
from sympy.abc import a,b,c

def phi(x,y):

    if y != 0 and y not in (a,b,c):
        return x / y
    elif y in (a,b,c):
        return x * y
    elif x == 0 and y == 0: