Konstantin Varik meule

A Few Useful Things to Know about Machine Learning

The paper presents some key lessons and "folk wisdom" that machine learning researchers and practitioners have learnt from experience and which are hard to find in textbooks.

1. Learning = Representation + Evaluation + Optimization

All machine learning algorithms have three components:

Representation for a learner is the set if classifiers/functions that can be possibly learnt. This set is called hypothesis space. If a function is not in hypothesis space, it can not be learnt.
Evaluation function tells how good the machine learning model is.
Optimisation is the method to search for the most optimal learning model.

Fun with vector maps

Quick TopoJSON vector tile demo map, derived from Nelson Minar's awesome work.

A goofy slippy map of various vector tile data sources.

Want to make maps like this? See my vector tiles tutorial.

Tested with Chrome. If the background tiles are pepto-bismol pink instead of gunmetal grey, your browser doesn't support -webkit-filter.

This example, using satirical data from The Onion, demonstrates how to wrap long axis labels to fit on multiple lines.

Update (jimkang): Added a data accessor function as a parameter to mbostock's wrap() so that it can fetch the text from the data source rather than using text().

The problem with text() is that it is built on textContent which "returns the text content of this node and its descendants." If a text element has several tspan descendants, it joins their contents without spaces. e.g. a text element with three tspans each containing 'why', 'are', and 'we' will give back 'whyarewe' through text(). The rest of wrap will proceed to treat that as one word, naturally.

This becomes an issue when wrap() is called multiple times on text elements that have already been processed by wrap().

	### JHW 2018
	import numpy as np
	import umap



	# This code from the excellent module at:
	# https://stackoverflow.com/questions/4643647/fast-prime-factorization-module

	import random

	import urllib
	import datetime as dt
	import pytz
	from PIL import Image
	import numpy as np
	import subprocess
	import socket
	import os

	dir = '/FULL/PATH/TO/SCRIPT/DIR'


	/**
	* Based on the mysql cluster
	* @link http://jonisalonen.com/2012/k-means-clustering-in-mysql/
	*/
	-- SELECT
	-- * FROM
	-- unnest(kmeans(array(SELECT
	-- (
	-- lat,

	From https://github.com/NelsonMinar/vector-river-map (Could be great for PLSSFinder reset, includes basemap and server configuration)

	We should set up an NGINX server if we reconfigure the Ubuntu Servers.

	Mapbox studio creates vector tiles using protocol buffer format. This is efficient, but also not supported by tilestache currently. We can, however, render JSONs from a postgres database and use the cartoCSS from mapbox for styling.

	If one would like to open a project created from a different machine, paths must be changed in the following files (using LCFRstyle4 and LCFRtiles for example, your names may vary):

	LCFRstyle4.tm2/project.yml
	LCFRstyle4.tm2/project.xml

	# put this in your .bash_profile
	if [ $ITERM_SESSION_ID ]; then
	export PROMPT_COMMAND='echo -ne "\033];${PWD##*/}\007"; ':"$PROMPT_COMMAND";
	fi

	# Piece-by-Piece Explanation:
	# the if condition makes sure we only screw with $PROMPT_COMMAND if we're in an iTerm environment
	# iTerm happens to give each session a unique $ITERM_SESSION_ID we can use, $ITERM_PROFILE is an option too
	# the $PROMPT_COMMAND environment variable is executed every time a command is run
	# see: ss64.com/bash/syntax-prompt.html

	function toWKT(layer) {
	var lng, lat, coords = [];
	if (layer instanceof L.Polygon \|\| layer instanceof L.Polyline) {
	var latlngs = layer.getLatLngs();
	for (var i = 0; i < latlngs.length; i++) {
	latlngs[i]
	coords.push(latlngs[i].lng + " " + latlngs[i].lat);
	if (i === 0) {
	lng = latlngs[i].lng;
	lat = latlngs[i].lat;

	#!/bin/bash
	#
	# Watch current directory (recursively) for file changes, and execute
	# a command when a file or directory is created, modified or deleted.
	#
	# Written by: Senko Rasic <senko.rasic@dobarkod.hr>
	#
	# Requires Linux, bash and inotifywait (from inotify-tools package).
	#
	# To avoid executing the command multiple times when a sequence of