Try it on paper first. A hint, the key idea and the worked solution are here when you need them.
A filter calls an email unwanted when its score is greater than its dividing line . It starts with . After each email it checks the user's label:
Five emails arrive in this order, as (score, label): , , , , .
Trace the filter. How many mistakes does it make, and what is at the end?
Machine learning is when a program has the ability to automatically adapt its own processes and/or data. It learns from examples (training data). An expert system's rules are written by people; a learning program adapts its own rules and data.
Have you tried it?
Work it out on paper first. It is fine to get stuck or make a mistake: that is how we learn. When you are ready, go through the solution one step at a time, and check each step against your own.
Plan. One email at a time: predict with the current , compare with the label, then change if needed.
Answer: mistakes, and at the end.
What it shows. The program automatically changes its own data, the value of , after each mistake. These five emails pull both ways, so it ends where it started: a few examples are not enough to learn a good line.
Log in to join the discussion.
Loading…