Limited Information Games: An Introduction to Deficiency
Brendan van Rooyen
NICTA SML SEMINARDATE: 2013-08-22
TIME: 11:15:00 - 12:15:00
LOCATION: NICTA - 7 London Circuit
CONTACT: JavaScript must be enabled to display this email address.
ABSTRACT:
In machine learning we are faced with several different types of data with which we can learn from. We are also faced with several different things we might wish to learn. For example in classification, from some training data I may wish to learn how to accurately predict the corresponding label from a given input. There are many different sorts of training data sets that one can use, ranging from standard supervised learning data sets to semi-supervised and many other more exotic creations. Understanding what sort of data is effective to learn from is key to making informed decisions about which data sets to acquire. There exists a large literature on this very question in a field known as the Comparison of Experiments. This talk aims to be a brief introduction to these ideas. There will be some theorems stated (without proof of course), some ideas introduced and time permitting an example of the sorts of problems these ideas can help solve.
