TODO

Some project ideas:



API for mldata.org
==================

Provide a programmatic interface to mldata, instead of the HTML webpage.

Implement the following interface for Data objects, which is heavily based on Google Docs API.
http://code.google.com/apis/documents/overview.html

We can provide the API under the url:
http://mldata.org/api/

First, we should allow users to access Data objects.

Discovery
---------
Retrieve Data lists that match specific tags. We already have this:
http://mldata.org/repository/data/?page=1&pp=all
http://mldata.org/repository/tags/data/Clustering/
But it would be good to return a json or xml blob for the API.

We could also support some "slicing", e.g. Data for multiclass classification
with between 5-10 classes, and less than 1000 examples. Currently, our database
design doesn't support this yet, but it may be good to keep in mind.

Revisions
---------
Review, download Data item's complete revision history.

Download
--------
Export documents in supported formats.
Each item is an ID, and so we can use this ID for retrieval.
http://mldata.org/repository/view/stockvalues/
http://mldata.org/repository/data/download/5102/
http://mldata.org/repository/data/download/xml/5102/

I suggested that we should use the slug instead of the id for the URL, but since
we have revisions, it might be good to keep the id. Then the discovery stage
should provide "groups" of revisions, etc.

Create/upload/copy documents
----------------------------
This functionality is currently done internally (see ml2h5), but it can be
pushed outside.

As a second step, we should allow editing of Data objects directly. This is more
efficient than downloading, editing, and uploading, if the data is large. But
perhaps we keep this for another day...
http://code.google.com/apis/spreadsheets/data/1.0/developers_guide_python.html




Integration with shogun http://www.shogun-toolbox.org/
=======================
The machine learning data set repository mldata.org provides data and
task downloads in a standardized format. Make shogun to natively be able
to read these data formats and interact with the mldata.org website,
i.e., by

1. read and write support for the hdf5 based data sets and tasks (see
http://mldata.org/about/hdf5/)

2. contribute features to mldata.org for command-line based
up/downloading of data,task,methods,solutions

3. contribute features to mloss.org for command-line based
up/downloading of methods and other community features

Note that the optional part of this task aims to help the machine
learning community portals for which python and django skills are needed
(both portal website's source code are also open source).





Create custom challenge view for Machine Learning challenges
============================================================

http://www.chalearn.org/challenges.html
Related to (private) custom view for Pascal VOC challenge by Lukasz Kidinski.



Create JSON API for data search
================================

Create a json api for data search. A draft for the specification could be:

GET /api/v1.0/search?q=QUERY&filter=FOO:BAR

200 OK
{
    'data': [
        {
         'author':  string,
         'date': datetime-string,
         'license': str,
         'format': [str],
         'rating': range(5),
         'views': int,
         'downloads': int,
         'views': int,
         'tags': [str],
        }
    ]
}