dlebech

<?php

# Public Domain CC0 license. https://creativecommons.org/publicdomain/zero/1.0/

# Install requirements:
# composer require ifsnop/mysqldump-php
# composer require google/cloud-storage

require_once __DIR__ . '/vendor/autoload.php';
require_once __DIR__ . '/config.php';

dlebech

# Public Domain CC0 license. https://creativecommons.org/publicdomain/zero/1.0/

# Prepare with: pip install fasttext
# Tested with Python 3.9

import urllib.request
import fasttext

# Download small model (917KB)
# Other options: https://fasttext.cc/docs/en/language-identification.html

dlebech

# Make * work in bash or zsh:
# zsh
setopt extendedglob
# bash
shopt -s extglob

# Resize all images (in-place) in a folder
# to 1200 pixels on the longest side
mogrify -resize 1200 *.jpg

dlebech

-- Public Domain CC0 license. https://creativecommons.org/publicdomain/zero/1.0/

-- Calculate the probability of k successes for n trials with probability of success k, 
-- using the binomial distribution.

-- Calculate the binomial coefficient using the "multiplicative formula"
CREATE OR REPLACE FUNCTION functions.binomial_coef(n INT64, k INT64) AS ((
  -- k!/(n!*(n-k)!)
  -- We're going to have a hard time doing factorials here,
  -- but based on the "multiplicative formula" in Wiki, it should be possible:

dlebech

# Public Domain CC0 license. https://creativecommons.org/publicdomain/zero/1.0/
#
# Download images from The Painting Dataset: https://www.robots.ox.ac.uk/~vgg/data/paintings/painting_dataset_2018.xlsx
# The image urls are outdaed in the Excel sheet but the painting urls are not,
# so this script re-crawls those images and downloads them locally.
# It works as of July 2020.
#
# Run this first with:
# $ scrapy runspider paintings_crawl.py -o paintings.json
# Images are stored in 'out/raw'

dlebech

# Public Domain CC0 license. https://creativecommons.org/publicdomain/zero/1.0/

# Run this file first, e.g.:
# $ scrapy runspider ft_extract.py -o members.json
#
# It will probably stop working if they change their urls for the contact list of course.
# Worked in Spring of 2019
import scrapy
import re
from urllib.parse import urlparse, urlunparse

dlebech

// Public Domain CC0 license. https://creativecommons.org/publicdomain/zero/1.0/
class Tokenizer {
  constructor(config = {}) {
    this.filters = config.filters || /[\\.,/#!$%^&*;:{}=\-_`~()]/g;
    this.lower = typeof config.lower === 'undefined' ? true : config.lower;

    // Primary indexing methods. Word to index and index to word.
    this.wordIndex = {};
    this.indexWord = {};

dlebech

# Matplotlib

# Creating a list of colors (e.g. for a bar chart)
# "Blues" is the colormap. It can be any colormap
# https://matplotlib.org/examples/color/colormaps_reference.html
colors = [matplotlib.colors.to_hex(c) for c in plt.cm.Blues(np.linspace(0, 1, len(some_dataframe.index)))]

# Globally adjusting DPI and figure size
matplotlib.rcParams['figure.dpi'] = 100
matplotlib.rcParams['figure.figsize'] = [6.0, 4.0]

dlebech

# Public Domain CC0 license. https://creativecommons.org/publicdomain/zero/1.0/
# Create a Keras embedding layer with an initial one-hot encoding by using identity initializer
import tensorflow as tf
import numpy as np

# Input sequence consisting of four features (e.g. words)
# Let's pretend this is "hello world hello everyone else"
# Where hello is then mapped to 1, world = 0, everyone = 2, else = 3, 
a = np.array([[1, 0, 1, 2, 3]])

dlebech

# Convert a unix timestamp in millisconds in a column of a CSV to a date 
cat thefile.csv | perl -MPOSIX -pe 's/(^\d+),/strftime("%F,", localtime($1\/1000))/ge'