Actions

Never Trust Any Published Algorithm

-


Event
MinneBar 10 (11 April 2015)
Panel
This is not a panel.
Summary
None
URL
Topic
Schedule
2015-04-11 09:15


Room: Discovery


Presenters



Certainly any algorithm that's been peer-reviewed multiple times must not have any obvious errors, right?

What about algorithms by leaders in the field, which come with proofs of correctness, and form the basis for tons of later research? Nothing of that stature could be flawed, could it?

But those are purely academic concerns, with no practical impact. It surely couldn't be the case that something as basic as a sorting algorithm, which was implemented multiple times and tested, fails to operate correctly?

In fact, I'll show examples of all three of these. Let's have a conversation about the ways in which algorithms fail--- and the ways to increase confidence that your algorithms and designs are correct.

blog comments powered by Disqus



MediaWiki spam blocked by CleanTalk.