linkdump - machine_learning

Restricted Boltzmann Machine, a Short Tutorial

nospam@example.com (Jochen Wersdörfer) — Thu, 30 Jul 2009 13:20:22 +0200

submitted by jb55
[link] [3 comments] [-->]

Tutorial with python source...

The $1M Netflix Grand Prize taken by BellKor’s Pragmatic Chaos?

nospam@example.com (Jochen Wersdörfer) — Sat, 27 Jun 2009 13:16:03 +0200

BellKor’s Pragmatic Chaos has broken the 10% barrier, a feat that may have won them the $1M Netflix prize. We’ll know for sure in 30 days.
[-->]

Congrats - impressive archievement...

Going Mini: Extreme Lightweight Spam Filters

nospam@example.com (Jochen Wersdörfer) — Thu, 25 Jun 2009 15:39:25 +0200

via Recent Google Publications (Atom) - (D. Sculley, Gordon V. Cormack):

Going Mini: Extreme Lightweight Spam Filters, D. Sculley, Gordon V. Cormack [-->]

Nice result. I wonder if it's really necessary to train only on the top features. Why not train on all features and then select the top weigths for a smaller model? If the goal is only to be able to store all models in memory, that should be enough.

Why I Don't Buy Clustering Axioms

nospam@example.com (Jochen Wersdörfer) — Sat, 20 Jun 2009 20:32:54 +0200

via natural language processing blog - (hal):

In NIPS 15, Jon Kleinberg presented some impossibility results for clustering. The idea is to specify three axioms that all clustering functions should obey and examine those axioms.

Let (X,d) be a metric ... [-->]

Clustering is a hard problem. I'm not sure if it's even possible to measure success. As Ernest Rutherford said:

All science is either physics or stamp collecting.

Maybe clustering is more on the stamp-collecting side...

The unreasonable effectiveness of data

nospam@example.com (Jochen Wersdörfer) — Sat, 20 Jun 2009 20:32:53 +0200

via vetta project - (Shane Legg):

We recently had a visitor to the Gatsby Unit talk about his work in reinforcement learning, in particular the use of planning and forward models to speed up the learning of difficult tasks. The substance of his talk was good, but that’s not what I want to talk about: it was the motivation he gave ... [-->]

From my own experience, i also think, that learning on a small scale is solved. And i also believe, that structured data alone cannot enable semantic interpretation. So, building up models from huge amounts of data seems to be the only option today. But still - we as humans don't have to read the whole web to learn about a language... something is missing there.

Dinosaur Comics - June 17th, 2009 - awesome fun times!

nospam@example.com (Jochen Wersdörfer) — Fri, 19 Jun 2009 23:07:27 +0200

via Delicious/csantos - (csantos):

In which T-REX solves all NLP problems ... [-->]

A little bit too close to truth, to be really funny.

Chain Conditional Random Fields: Implementation and Design Issues

nospam@example.com (Jochen Wersdörfer) — Fri, 19 Jun 2009 21:47:45 +0200

via LingPipe Blog - (lingpipe):

I’m starting to implement conditional random fields (CRF) for LingPipe. (See, I really don’t hate CRFs.) If you don’t know CRFs, here are three good tutorials: [-->]

Really good tutorials, enjoyed reading them.

Enabling Exploration Through Text Analytics

nospam@example.com (Jochen Wersdörfer) — Tue, 16 Jun 2009 22:26:54 +0200

via Science for SEO - (CJ):

I’m a huge fan of Daniel Tunkelang and his blog “The noisy channel“. Recently he spoke at the Text Analytics Summit and was kind enough to share his slides with us all. Check out the other available prese ... [-->]

I think exploratory search is still an unsolved problem, but nice slides anyway...

Least squares fit of a surface in Python (with ridiculous application)

nospam@example.com (Jochen Wersdörfer) — Tue, 16 Jun 2009 22:26:49 +0200

via Hacker News:

Comments [-->]

Bad floor, but nice math.