Jump to content

User:Graeme E. Smith/Collections/Model Series/Datamining/Sampling Techniques

From Wikibooks, open books for an open world

Sampling Techniques

[edit | edit source]


Let us use a fishing metaphor for sampling techniques. You know that your mind is conspiring with itself to hold information that is interesting and useful, away from your declarative memory. The problem is that there is no hook you can use to find the information. It is a little like knowing that there is fish in a creek but no idea where to drop your line to hook one. Anyone who has just taken their line and plopped it down in the first available spot in the creek, knows that they have to be especially lucky to actually catch anything. So good fishermen know that the trick is to have a strategy of where to put your hook, and depending on how good your strategy is, you will have better or worse luck. If all you want is to sit quietly for a while, you can take your time, but if what you want is to catch a fish, strategy is the game to play.

Casting

[edit | edit source]


One technique you might want to try is casting. Essentially Casting is a sampling technique where you tease the surface of your thoughts with fly light taps, and see if there is anything waiting to bite. Once you find a place where fish are biting you can spend more time casting in that area to flesh out the knowledge you have found. The idea is to flick different areas of your thoughts each time you cast, until you get a bite. What you need is an idea of the boundaries of the area you want to study, and you randomly pick areas within that area to cast to. A good fisherman can study the areas where bites happen so that they will know where to try casting during their sampling next time.
What is casting doing? Well, in actuality the mind stores everything in the implicit memory areas, and what you are really doing is rehearsing a clump, and looking to see if the Quale it addresses has changed from your internal representation of it. If the Quale is different, you want to explore the difference to see if you have learned something you didn't know before.
So by randomly sampling your implicit memory by picking explicit clumps to rehearse, you are sampling the knowledge you know to see how it is changing. Once you know that your knowledge has changed, you can concentrate on the way that it has changed, and build up a picture of the sub-rosa knowledge that is hiding in your explicit memory.

Trolling

[edit | edit source]


Another technique that fishermen use, that you might want to try is a sampling technique called Trolling. This technique is especially good, in situations where you are learning to flesh out an area of knowledge, and don't quite know enough about it to actually cast around in it. I want to separate the idea of trolling in datamining from the idea of trolling in the internet. One has to do with dragging a lure past a number of hiding places for knowledge, while the other has to do with a type of posting that is more interested in the number of responses it can generate than in the content. There are simularities in technique, but we are interested in content, not attention.
Many people troll their intuitions without realizing that they are doing so. However once you understand what is involved, you can intentionally troll, and in doing so learn more about your own mind. Trolling is a Synthetic process rather than a random access process, and so it tells you something about the completeness of the information you know, when you lose your thread, you have reached a place where perhaps your knowledge is not adequate to sustain synthesis. This can guide your acceleration processes.
The idea is to build or synthesize an explanation of your knowledge. Academics do this by writing papers on their subject of interest. Wikians often do it by building a wiki-site of one sort or another around their knowledge. It is important to realize that it may take you 2 years or more for your knowledge of a subject to become deep enough that you can write about it normally but with trolling you can begin writing much quicker, because you create opportunities to transfer knowledge that is implicit into the declarative memory where you can express it in writing.
A good trolling technique is to sit down and start to write everything you know about a subject, you will find as you go along, that many things you thought you knew have changed slightly with more experience, and as a result, the topic itself becomes more interesting and informative than you would have thought it would be. This book started when I tried to sit down and write everything I knew about intuition. It is often best to structure your writing so that you write a paragraph, or a page or three for every topic you think you know something about within the larger topic. One way of doing this is simply to lay out an outline of your topic, and fill it in with data. How you use the information once you have written it, is of course up to you.