First Approach to Neural Networks

Thursday,March 21 ,2008

If you have learned enough about neural networks, and you realize that the implementation is quite simple, now suppose we already have our completely new software designed to recognize faces of people and the main task of this software is to distinguish gender: female, male. We can do this task easily taking into account the social order. Women use makeup and long hair usually factions are more refined, the man usually uses short hair and factions tend to be more rough, an artificial neural network can also make the task of making a reconnaissance image. Does our new software can accomplish the task?

If we implemented the architecture of the neural network and the network is connected to supplements such as camera, digital filters, yet our neural network does not know anything, it's like a baby who barely the knows the world, the neural network must be trained to be able to recognize what we want to recognize.

To solve the problem, what can we do?

A training system must be implemented using an training algorithm.
Now suppose we develop new software capable of rebuilding lost images. This gives us one more problem, we should implement another training system for our other new software.

The solution that I have found in both cases is the implementation of an universal trainer,that will give us the great advantage of to train our neural networks often necessary or modify network parameters. Without the need to go directly to modify the code, avoiding waste of time and money.

For months I have been develop this project, usually it is a graphical user interface that serves as a mediator with the various neural network models, in other words, it supports the different neural networks architectures, and not just in the design part, but also in the programming part, a library makes iteration with our program in our case, the face recognition software, thus our program alone is responsible for providing the inputs and to receive outputs without the need to interact directly with the neural network internal data.

The breadth of this application is not limited to software, also can be applied in hardware implementation, where generally weights of the neural network should be fixed and where develop an internal training system is too complicated.

Currently the development of this project is in place and will be ready shortly, and I hope more than anything, that it will be helpful for developers of new technologies.



Category Neural network trainer project

Comments (0)

You may use html tags <a href="url">Your link</a>,<b></b>,<ul><li>.

Name:
Email(Optional):
URL(Optional):
Comment:
Are you human? if you are answer this simple question:
What color was Napolion's white horse?