A random walk

A talk all about random numbers and randomness.

Of course, the slides will come in random order - some maybe not at all, some twice or more, maybe there'll just be one that comes again and again and again.

I hope to learn/uncover things like:

using predictable random numbers to give the illusion of infinity
bloom filters and other statistical data structures
markov chains for text generation
monte-carlo methods
reducing collisions without locking
guessing by sampling
how ruby's random works
simulated annealing
random deployment

I don't know all the things I'll need to learn for this talk, so please suggest topics.

Proposed by: Ben Griffiths updated 8 months ago

Suggestions

Graham Ashton suggests: 8 months ago

Markov chains, monte-carlo and simulated annealing are really interesting (and frequently useful). +1
Matt Smith suggests: 8 months ago

Most interested in random deployment and learning about how Ruby's random works.

Hope the slides come up!
Joe Corcoran suggests: 8 months ago

I agree with the comment below about using Markov chains to simulate user behaviour – I'd love to know more about that.
Tom Taylor suggests: 8 months ago

+1 for bloom filters, simulated annealing, genetic algorithms and all that guff.
Murray Steele suggests: 8 months ago

I love the conceit of a random set of slides for a talk about randomness.

The only thing I'd like to hear about would be rather dull, but something that I encounter a lot. Often a client will ask for some element of a page to be random. I'd like to know how best to a develop that (ORDER BY RAND() LIMIT 1 can't be the best way) and how to test it (especially if I've pushed the randomness to the database).
Ben Griffiths suggests: 8 months ago

Harry, Tom: I'm not a computer scientist, just a hacker, so I'll pitch it at the level that I think I would understand. I don't have Maths A-level so it probably won't include all that x and y stuff.
Harry Rickards suggests: 8 months ago

I think this would be really fascinating, but as Tom said: what sort of previous knowledge would be required? It's understandable if we will need some sort of foreknowledge to understand, but if someone could link to some webpages that will bring beginners (maybe assume Maths A-Level knowledge?) up to scratch for before the talk, that would be great.

(Halfway through writing this, realised you're the same @beng who sent me the Magic Machine book - it's awesome, many thanks for it :D)
Tim Cowlishaw suggests: 8 months ago

Another practical application that just occurred to me that might be interesting if you have the time and wherewithal to fit it in - using markov chains to simulate user behaviour for load testing a la http://ieeexplore.ieee.org/iel5/5581107/5581301/05581372.pdf
Tom Stuart suggests: 8 months ago

As a self-trained code monkey with no formal computer science or mathematical training, this is fascinating to me but I'm a little concerned it would go over my head. At what level are you thinking of pitching this, Ben?
James Adam suggests: 8 months ago

I'm a sucker for the random slide order thing. Do you imagine any vaguely-narrative thread linking the topics, or will it be an eclectic epileptic frenzy?
Richard Drake suggests: 8 months ago

I've been learning about spatiotemporal chaos recently and would be disappointed if this session didn't produce it, if only as an emergent property. Ref: http://crossgroup.caltech.edu/STChaos/index.html - sorry about the Java.
Andrew Nesbitt suggests: 8 months ago

I can't wait to see how the slides come out for the talk!
Anthony Green suggests: 8 months ago

I have no idea what any of these terms mean so a positive +1 from me for this talk. Should we do some background reading ?
Tom Stuart suggests: 8 months ago

Love this idea. I'm particularly intrigued by "random deployment".
Abdel A Saleh suggests: 8 months ago

Excitings, looking forward to this, especially: - Bloom filters - Predictable random numbers - guessing by sampling
Tom Lea suggests: 8 months ago

Very wide ranging, perticularly interested in monte-carlo methods, reducing collisions without locking, and simulated annealing.

Predictable random feels like a talk in itself, but let's see how the dice land.

Random deployment + Chaos monkey (using randomness to add a certainty that things will break, and thus determining dev practices) could be interesting.
Levent Ali suggests: 8 months ago

I have to deal with a lot of unverified data sources that claim to be similar types of data. Anything that can help us identify sources that could be related and/or are unrelated would be of tremendous use as well as being very interesting.

I think James Mead somewhat expressed what I think I'm trying to.
Tim Cowlishaw suggests: 8 months ago

This sounds very interesting indeed. It might also be interesting to touch on Naive Bayesian Classification and Bayes nets - both interesting and useful probabilistic techniques that would fit well alongside this.
Chris O'Sullivan suggests: 8 months ago

Random deployment? Like, your app deploying just...whenever?
James Mead suggests: 8 months ago

I'd be interested to learn about useful mathematical/statistical tests for randomness.
Paul Battley suggests: 8 months ago

It might be covered by how Ruby's random works, but a brief explanation of entropy pools (and how they collect random data for the pool) could be illuminating.