PAC learning

Add to your list(s) Download to your calendar using vCal

• Peter Rugg, Churchill College
• Wednesday 12 October 2016, 19:00-19:40
• Wolfson Hall, Churchill College.

If you have a question about this talk, please contact Jasper Lee.

Machine learning allows computers to solve problems without being explicitly programmed with the solution. However, what sorts of problems can be learned? What does it even mean to learn a problem? The Probably Approximately Correct (PAC) machine learning framework addresses these questions, specifying worst case error bounds before a problem can be said to be learnable. In this talk, we will see a formulation of the supervised binary classification problem, followed by a definition of PAC learning. We will then try to find a way of determining which problems are PAC learnable, and which are not.

This talk is part of the Churchill CompSci Talks series.

This talk is included in these lists:

• Cambridge Centre for Data-Driven Discovery (C2D3)
• Cambridge talks
• Chris Davis' list
• Churchill CompSci Talks
• Interested Talks
• Trust & Technology Initiative - interesting events
• Wolfson Hall, Churchill College
• bld31
• computer science
• ndk22's list
• ob366-ai4er
• rp587
• se393's list

Note that ex-directory lists are not shown.