Report

Original Plan

An investigation into neural networks or other systems which copy some of the ways the brain works. In this investigation I will try to cover:

• what are neural nets

• a brief history of their development

• what they can do/what are they used for

• a comparison with normal computer systems

• making a small neural network simulation

Full Page

Log

Evaluation

The objective I set myself was, To learn as much as is feasible about neural nets and to present the information so that other people will be able to understand it.

This is a very general statement but I feel that I have carried it out. The resurge in interest in neural nets is relatively new and this means there is not a full spectrum of information about it. At one end there is highly technical information which would be very difficult to understand and at the other there is the fact that they model the brain. There is little middle ground which gives detail but is not impossible to understand.

This is the area which I feel my investigation fits into. In section 1 I laid out the basic aims of the investigation and gave a problem of the computer industry and the fact that neural nets may be able to solve this. Section 2 has a history of the computer industry which introduces new ideas for someone not knowledgeable about computers, it also includes information about the brain. This section acts as a backdrop against which neural nets can be presented. The workings of neural nets is described in section 3, it includes much more than a basic explanation of them but does not go into detail about why they work. This is about as much detail as I could easily understand after that it becomes necessary to use tables of number and formulae. Section 4 was relatively easy as all that was required was to sort out the examples and reword and shorten them so they were suitable for my investigation. After those four sections there is still a lot of information that I had not touched and ideas which I could have covered. Section 5 introduces something that the reader can think about that of computer and minds and I give my opinion about neural nets which the reader can either agree or disagree with.

My investigation was meant to be understandable and to that end I added section which contains a dictionary of all words that I though may cause problems to some readers. To make it possible for readers to go back to check something all the words are given page numbers as are the names of all the people mentioned in the investigation.

In my full page outline of my project things are a little different from the final version. One main part that has gone is the Timeline which was meant to contain a list of dates important to neural nets. In the final version I left this out because it did not seem to fit in with the other appendices. The dictionary and the Who's Who both allowed the reader to look up things they may not know about or to find out where in the project they were first told about them. the Timeline would have been of far less help because, in the investigation, I did not refer back to things via the dates. Therefore I decided to leave the Timeline out. If I was doing the investigation again I might include it if I changed they way in which I referred to previous information.

There is also less comparison than the full page suggests. This is just due to the way in which I wrote the investigation. Most of the information I intended to put in has been included but I have not spent much time comparing information, leaving that for the reader to do if they are so inclined. This also means the investigation is less bias as fewer of my opinions have therefore been allowed to come in.

One major area that I planned to cover but didn't was designing a small simulated neural nets and giving details of its construction and performance. This was far to grandiose an idea to carry out which I learned after I had read some of my reference material. Also I realised that it would be very hard to show to the reader any useful information as what would probably have come out is a large list of numbers which showed its connections and performance. If I did this project again I would not include this part unless I was much more knowledgeable about neural nets and my project as much longer and was written for an experienced reader.