Is pattern recognition the same as unsupervised learning? Is machine learning the same as supervised learning?

Question

Firstly, here is the definition of a well-posed learning problem:

A computer program is said to learn from experience E with respect to some task T and some performance measure P, if its performance on T, as measured by P, improves with experience E.

My machine learning professor claims that supervised learning is machine learning and unsupervised learning is pattern recognition. Here are his definitions of machine learning and pattern recognition:

Machine Learning is learning from experience. It’s also called supervised learning. E consists of features and labels, and P and T are well-defined.

And

Pattern Recognition is finding patterns without experience. It’s also called unsupervised learning. E consists of only features, and P and T are defined in much broader terms of finding ‘interesting patterns'.

If I cross reference this with other sources on the internet, I get different definitions. They say that machine learning involves both supervised and unsupervised learning. Likewise, pattern recognition involves both supervised and unsupervised learning. Some say that pattern recognition is machine learning. Some even say the opposite, that machine learning is a form of pattern recognition.

So is my professor right, or is the internet right, or is the answer somewhere in between? Can we answer this question definitively?

Answer 1

Nice topic. I think it depends on how the "learning" process by a machine is defined, so if we assume that a machine can "autonomously" learn from experience (i.e. new data as it becomes available every X time), why not to include also semi-supervised and unsupervised learning styles? The more data the machine has about the system being modeled, the more info to find more precise patterns which correctly describe the system, for all kind of learning styles.

As an example I made to explain how the one-class support vector machine works, we could see that, although an unsupervised learning process does not have a ground truth (i.e. labels) or a semi-supervised learning has only the "normality" label, the machine is able to better describe the system as it receives more data and as some hyperparameters are adjusted, if you test your model with known points wich should be outside the learnt cluster (where an anomaly score could be also a metric in this case).