daily #4

Today, one of lab members told me about genetic programming. 

I was really happy to know the theory. It was like I finally found something that draw me in. 

It represents what I believe, in the way I believe. I would like to know the more. 

However, what he said after that was surprising. 

"I see you like theories like this, but I don't like it. No brilliant algorithms are involved, slow, and most of all, there is no answer"

Number is very powerful. We can do anything with number using properties and theorems that people have improved so far, if we can clearly model the real world to the numbers. Suppose we have 48 apples, and 3 man, and our job is to distribute applies equally to 3 people so that no one gets mad. What makes this problem extremely simple is the number. The problem above is simple 48 / 3 problem because we model forty-eight apples into the number 48, three man into the number 3, ignoring every properties of them, but just quantizing it. Now it's going easy, operations such as subtraction, addition, and multiplication can be used. 

However, the hardest part is to model them as number. What if we have a apple that is very very small so that it's hard to be counted as one apple, or what if we have a mutation Siamese twins apple. Then someone who gets the very small apple will get mad, and those who do not get the Siamese twins apple will also get mad. At the bridge between the theory and practice, we got to do something, following the rule by counting them as one apple, or making different rule to apply. It seems the number is not perfect enough to mimic the imperfection of the real world, rather than the world is not perfect enough to mimic the perfection of the number. What do you think? what is not perfect? 

I think I like the theories that are able to represent imperfection. The algorithms that can find the perfect answer according to the perfect process is not attractive to me. I might want to see the trial and error in the way we try to solve a problem, and I like the algorithms that contains that kind of philosophy. The reason why machine learning is attractive at the first time I learned is because of this. The philosophy of machine learning is that we don't explicitly code the process of program.