Projects that we are considering working on.
-
Email classification to sort an inbox into "important" and normal mail without using Gmail; possibly extend it to filter spam more reliably; understand the (theoretical and practical) differences between Bayesian (usually employed for that purpose) and neural network algorithms (which might allow to invest more CPU time to get better precision)
-
Fix typos in texts written on mobile phone touch screen (according to most likely word to fit, considering sentence context well as proximity of the mis-typed key to the correct one)
-
Teach a neural network to draw logic conclusions. Similar to this?: Neural Programmer-Interpreters / arXiv / HN
- see whether AI can help find 'patterns' in genome data, or music
- evaluate data from commoncrawl.org
- Image recognition of foxes versus cats versus humans in video streams (perhaps note Andrew Ng's research?)
- Enter one of the competitions from [Kaggle] (http://www.kaggle.com) or similar