### Andrew Walkingshaw

Shared publicly -Why I'm suspicious of "data as physical thing" metaphors, and why data mining has more in common with user experience design than you'd probably expect.

1

1

Add a comment...

Andrew Walkingshaw

Attended University of Cambridge

Lives in San Francisco, CA 94103

243 followers|3,140 views

AboutPosts

Why I'm suspicious of "data as physical thing" metaphors, and why data mining has more in common with user experience design than you'd probably expect.

1

1

Add a comment...

So you can share photos and video, but not audio? -1, Google.

1

8 comments

A part of me thinks it's way easier with something like this – Google+ accounts are tied to real identities, so that makes them more valuable. Stick up a big banner saying that if a record label busts you, your account'll get shut down; post your own podcasts and music. Job done...

Add a comment...

In their circles

147 people

So here's something I've been thinking about a little bit which isn't really developed enough for my blog yet.

I'm all for more design thinking, basically, everywhere. Definitely in mainstream Internet services – things like this and Facebook, but equally the bane-of-our-existence Intranet apps we all wind up using. However, much of the material that good services are built of is pretty advanced, mathematically. It's hard to have a conversation about designing data-heavy services if you don't even have a shared language.

There's no simple solution to this, but one part of the answer, I suspect: I think the world needs "maths for Web people" course. A Project Euler thing, an O'Reilly thing, Khan Academy style lectures, a night-school course; I'm not quite sure what the format is. But something like that. The point wouldn't be to get someone to the point where they're deriving formulae by hand or anything, just to give a flavour of what's possible - enough to follow along, at least, and to reason about what approaches might be workable.

What would go in it, though? At a very high level, I'd come up with something like:

Elementary calculus

Dimensional analysis

Common statistical distributions

Complex numbers and the complex plane

Vector spaces and operations

Symmetry and symmetry groups

Matrix algebra (and the linear groups)

Elementary graph theory (up to, say, adjacency matrices and Travelling Salesman)

What I'm getting at is that once you've got all that, you've got enough maths to follow the principles behind machine learning, and it's not like we're going to have less of that any time soon.

Anyway: what am I missing? (And what's in here that shouldn't be?)

I'm all for more design thinking, basically, everywhere. Definitely in mainstream Internet services – things like this and Facebook, but equally the bane-of-our-existence Intranet apps we all wind up using. However, much of the material that good services are built of is pretty advanced, mathematically. It's hard to have a conversation about designing data-heavy services if you don't even have a shared language.

There's no simple solution to this, but one part of the answer, I suspect: I think the world needs "maths for Web people" course. A Project Euler thing, an O'Reilly thing, Khan Academy style lectures, a night-school course; I'm not quite sure what the format is. But something like that. The point wouldn't be to get someone to the point where they're deriving formulae by hand or anything, just to give a flavour of what's possible - enough to follow along, at least, and to reason about what approaches might be workable.

What would go in it, though? At a very high level, I'd come up with something like:

Elementary calculus

Dimensional analysis

Common statistical distributions

Complex numbers and the complex plane

Vector spaces and operations

Symmetry and symmetry groups

Matrix algebra (and the linear groups)

Elementary graph theory (up to, say, adjacency matrices and Travelling Salesman)

What I'm getting at is that once you've got all that, you've got enough maths to follow the principles behind machine learning, and it's not like we're going to have less of that any time soon.

Anyway: what am I missing? (And what's in here that shouldn't be?)

1

2

3 comments

I think these topics are a good introduction, but adding some more applied themes would be good to add colour emphasis. I'm thinking of basic machine learning techniques such as K-means, SVD, clustering, decision trees - those sorts of things.

Introducing concepts around optimisation and compression would be super useful for a lot of people, too.

Introducing concepts around optimisation and compression would be super useful for a lot of people, too.

Add a comment...

People

In their circles

147 people

Education

- University of CambridgePhD, Earth Sciences, 2002 - 2007
- University of CambridgeMSci, Mineral Physics, 1998 - 2002

Story

Tagline

Curmudgeonly data scientist.

Places

Currently

San Francisco, CA 94103

Previously

London N5, United Kingdom - Cambridge CB5, United Kingdom