Using this Coursera's course about machine learning, I learnt some things about supervised and unsupervised algorithms.

And from Wikipedia I've read that machine learning is "to get better performance/P at tasks/T with each task being done, that is experience/E".

Now, in those algorithms what I see is that they give it a set of data (either pre-labeled for supervised algorithms, or not for the other one), and the algorithms try to figure out the data and generate some output.

But I truly fail to find out WHERE and HOW in the process those algorithms BECOME BETTER with each new dataset they receive.

I really appreciate any help here. Do I miss something about machine learning? Without the getting-better-with-each-new-dataset part, how are these algorithms different from normal programs we right with a huge amount of conditional clauses and loops?

Using this Coursera's course about machine learning, I learned some things about supervised and unsupervised algorithms.

And from Wikipedia I've read that machine learning is "to get better performance/P at tasks/T with each task being done, that is experience/E".

Now, in those algorithms what I see is that they give it a set of data (either pre-labeled for supervised algorithms, or not for the other one) and the algorithms try to figure out the data and generate some output.

But I truly fail to find out WHERE and HOW in the process those algorithms BECOME BETTER with each new dataset they receive.

I really appreciate any help here. Did I miss something about machine learning? Without the getting-better-with-each-new-dataset part, how are these algorithms different from normal programs we write with a huge amount of conditional clauses and loops?