Wednesday, September 29, 2004

Neural networks

I just read an article about the up bringing of children. There is a discussion about the two extremes, caning and reasoning, and also some mixtures of the two. I might have mentioned before, another article I read some months back stating that it might not be so much dependent on the method, but the consistency of the rule or standard that underlines the method. For example, when on different occasions, a child acts in a particular desirable or undesirable way and he/she receive different reactions or responses from the parents, the child will be confused, find contradictions and inconsistencies.

Actually, many computer scientists and engineers should be very familiar with this phenomenon. For a number of decades, artificial neural networks (ANN) were implemented using mathematical models. On class of these ANN needs to be trained in order to perform in an acceptable way. The training effectiveness depends on the training data. If there were substantial portion of the training data has error, the ANN will probably not able to fulfill in role.

Now, we have to know that ANN is an artificial version of the neural network that scientists found in brains. I agree that brain is not the whole of a human being and therefore let alone the neural network in our brain. However, if the neural network in a child brain is fed with wrong training data by giving them different or even contradictory reactions and responses to the same behavior, then the that neural network will not be trained properly. This would result in the child behaving in a way that we label as undesirable.

So, may be some engineers and some psychiatrists should start working together for a better parental guideline in the up bringing of children.

No comments:

Related Posts with Thumbnails