Joseph Turian’s just created a site
which describes itself as a place “where data geeks ask and answer questions on machine learning, natural language processing, artificial intelligence, text analysis, information retrieval, search, data mining, statistical modeling, and data visualization!”
It’s modeled pretty directly on the hugely popular Stack Overflow, which bills itself as a “collaboratively edited question and answer site for programmers – regardless of platform or language.”
In case you’re too lazy to look yourself, questions at the top of Training Examples right now are
- how to discovery emerging topics in customer contacts?
- How do I test a method (murmurhash) for generating random numbers?
- Finding related users based on user data
- Algorithm to discover features from raw data?
- When is low-dimensional dense representations better than high-dimensional and sparse? And vice-versa?
- What are the state of the art optimization algorithms for problems with numerous local minima?
Par for the course, it’s a mix of wildly general (non-convex optimization) and reasonably specific (testing a random number generator) questions.
Joseph sites the following motivators for using Training Examples:
- Communicate with experts outside of your lab.
- Answer a question once publicly, instead of potentially many times over email.
- Share knowledge you have that might not be reflected in your publication record.
- Find new collaborators.
Those are also good reasons for starting a blog or commenting on one!
The site requires registration for posting or tracking questions, but not for reading questions or answers. I have gazillions of passwords and a regular process for managing them, but it’s still enough of a pain that I don’t just register at every site I visit. Of course, with something like this, you need registration, or it’ll be nothing but spam in ten minutes flat. It’s a constant battle on the blog, but I don’t want to moderate or require registration.