Finding Relevant Information
By ai-depot | June 30, 2002
Motivation
Saving time
Time is money, and reading background material will beyond any doubt allow you to start tackling the problem very quickly. This is true if you directly apply existing solutions, and even if you decide to adapt them to your needs. Most of the tricky issues will have been discussed at length already, so all you’ll be left with is the software design and implementation.
Re-inventing the wheel
You won’t get any prizes for doing what’s done before. In industry, you’ll have trouble marketing your product if the core technology is exactly what your competitor has been using for the past decade. In academic research, you’ll no doubt get a few laughs for having spent 3 years of your life working on something that was solved in the 1960’s. This has been seen before, notably for the development of the back-propagation algorithm, which was solved about 4 times independently.
What’s the name of that guy that walked on the moon second?
Learning from experience
Looking into existing approaches will give you a feel for the entire field, via the numerous success-stories and miserable failures. This provides you with valuable knowledge in order to avoid duplicating similar mistakes. You will also be able to pick out the methodology, giving you the possibility to taylor it to your own needs.
Definition and solution
Even if you are setting out to solve the problem in your own way, looking at previous research will outline how the problem was defined (even before it was solved). This definition can often be re-used for different solutions, and is in fact often the biggest step towards solving the problem. Your ability to re-use the same definition will permit your solution to cover a wider range of possible applications.
Benefiting from momentum
Research in a particular area is almost always heading in one specific direction, towards a goal that has been set out by initial research. In tackling your problem, you may often benefit from using such past approaches to enhance your own contribution to the field. You will also receive more appreciation from fellow researchers and programmers if your solution is useful to them!
Abstracting existing solutions
Quite often, initial attempts at solving problems are very rigid, defined by sets of rules developed by the author. When taken from a higher point of view, these can often be generalised and abstracted, allowing a more elegant solution. It is also commonplace to get the computers to learn the behaviour automatically, rather than define them explicitly. Obviously, you can only this can only achieve this if you’ve read the original papers!
Combining multiple fields
Many people claim “My research has never been done before, there’s nothing I can read.” That’s likely to be true about 0.0001% of the time — a number rounded upwards for convenience. No matter what you do, there is always one field of computer science or A.I. that has tackled a subset of your problem. It’s just a matter of finding it! No doubt every single aspect of it may be covered by a different field, which will leave you with the most research. The project itself, however, will be all the more interesting!
Tags: none
Category: essay |