|
There are many useful resources on the web about Artificial Neural Networks that can give some insights to interested individuals. Here are a few:
|
Artificial Neural Networks
Artificial Neural Networks are mathematical models that map input-output data in a similar way that a human brain would do. There are several different types of ANNs, but most of them share some common characteristics, such as they process through layers of parallel connected neurons, that have highly complex and nonlinear mechanisms, which were inspired by the biological neurons in the human brain. One of the most important principles of ANNs is they learn and remember the data and they have good generalization capabilities, so they are very good in responding to new data that they never saw before. Due to this characteristic, they are widely used in pattern recognition, machine learning and artificial intelligence applications as a nonlinear classifier.
Learning in ANNs can be categorized into 2 main groups:
- Supervised learning: Then network has a teacher who knows the right output and gives a learning performance feedback to the system.
- Unsupervised learning: The network decides on its own for what the output will be and categorizes the inputs accordingly
The most commonly used ANNs are single and multilayer perceptrons (backpropagation), Radial-Basis Function Networks (RBF) and Self Organizing Maps (SOM).
The application areas of Neural Networks are the same as Pattern Recognition. Some of them are:
- Character recognition
- Automated visual inspection
- Medical diagnosis
- Face recognition
- Credit card fraud
- Fingerprint Identification
- Remote sensing
- Motion analysis
- Bankruptcy prediction
|
