Comments for imagine27

Comment on O-notation considered harmful (use Analytic Combinatorics instead) by Evan Plaice

Evan Plaice — Tue, 12 Feb 2013 07:20:50 +0000

Unless the majority of the data is already sorted and it’s just a matter of sorting the new entries into the pre-existing structure. Then insertion sort works a lot faster than quick sort.

Try it for yourself: http://www.sorting-algorithms.com/

Case in point, algorithm complexity analysis is heavily context sensitive, regardless of what established theory claims.

I’m currently taking Robert’s Algorithms 1 course on Coursera but I’m looking forward to seeing what he has to say about combinatorics once I work my way up to it.

Comment on O-notation considered harmful (use Analytic Combinatorics instead) by The big-O notation is a teaching tool

The big-O notation is a teaching tool — Tue, 12 Feb 2013 01:44:08 +0000

[...] reading: O-notation considered harmful, “In the long [...]

Comment on O-notation considered harmful (use Analytic Combinatorics instead) by scott

scott — Mon, 11 Feb 2013 20:12:30 +0000

It’s not about speed, it’s about scalability.

Comment on O-notation considered harmful (use Analytic Combinatorics instead) by Tobias Florek

Tobias Florek — Mon, 11 Feb 2013 09:35:24 +0000

you will appreciate big-O notation, when you have to handle user input that can (intentionally) trigger worst-case behavior.

Comment on O-notation considered harmful (use Analytic Combinatorics instead) by Henk Poley

Henk Poley — Mon, 11 Feb 2013 08:59:43 +0000

With all the sorting going on here, I wonder why nobody has mentioned timsort.

Comment on O-notation considered harmful (use Analytic Combinatorics instead) by Steven

Steven — Mon, 11 Feb 2013 03:18:27 +0000

Quick sort is often used over merge sort not because there’s some magic gap between theory and practice that somehow makes an O(n^2) algorithm faster than an O(n log n) algorithm, but because we can easily show that quick sort runs in O(n log n) time for most inputs, and that randomised quick sort runs in expect O(n log n) time with high probability, and quick sort usually has a better constant factor on the machines we use.

A quick sort with pivot selection that can be easily exploited to make the algorithm run in O(n^2) is quite dangerous, and has been used as a DOS attack vector in the past.

You’re missing the point of this type of analysis entirely, and being extremely condescending while doing so.

Comment on O-notation considered harmful (use Analytic Combinatorics instead) by Arthur

Arthur — Mon, 11 Feb 2013 02:48:01 +0000

So many grammatical errors… my eyes…

TL;DR: Sedgewick has a new book out. (You can get it for free here: http://algo.inria.fr/flajolet/Publications/book.pdf)
OP has apparently failed numerous job interviews due to erroneously claiming “Quicksort is better than mergesort/heapsort/introsort,” and refusing to acknowledge asymptotic performance considerations. OP takes Sedgewick’s book as validation of his position. OP is still wrong.

Comment on O-notation considered harmful (use Analytic Combinatorics instead) by Jon Forrest

Jon Forrest — Mon, 11 Feb 2013 01:32:18 +0000

“and it’s misuse” should be “and its misuse”

Comment on O-notation considered harmful (use Analytic Combinatorics instead) by rrenaud

rrenaud — Mon, 11 Feb 2013 00:56:44 +0000

Google benchmarks its search system against historical logs, fighting against average case mostly.

Algs that use unobserved randomness can still do fine against adversaries.

Comment on O-notation considered harmful (use Analytic Combinatorics instead) by Arash Partow

Arash Partow — Sun, 10 Feb 2013 22:43:53 +0000

Take the Quick vs Merge sorts example, understanding the differences (in an exact nature), and reasoning about these difference (you need the tools), coupled together with questions that expose the mechanical in\efficiencies (eg: when is quick sort better than bubble sort?) of the implementations, allows one to make the leaps on intuition that lead to things like: Introsort.

So don’t pooh-pooh such things…. It is true that the various complexity notations are initially difficult to get one’s head around (especially if you don’t come from a deep math/functional analysis background), but they are good yard sticks for reasoning and direction purposes.

That said nothing beats empirical analysis, but any tool that reduces the effort is worth learning.