The first line of section 2.7.3 in Mitchell's Machine Learning is:
"A Learner that makes no prior assumptions regarding the identity of the target concept has no rational basis for classifying any unseen instances."
Why is it that a machine learning algorithm needs a bias? Can someone please help me understand this, with examples perhaps?
Inductive bias means all assumptions your learning approach assumes for generalising to unseen data. What we do in machine learning is induction which means we don't have rules. By seeing data, we make rules. For example, in linear regression you may consider that the output is linear with respect to inputs, or it is polynomial. Another example can be this that in women gesture recognition, you don't care about moustaches because you know they do not have that. You can also take a look at here.
