Perceptron is one type of ANN single-layer. Perceptron was first introduced by Frank Rosenblatt, and contained the training algorithms used to build the ANN model. The simplest perceptron uses a single processing neuron. Since it only uses one processing neuron, Perceptron can only classify two classes. If you want to classify more than two classes, we usually use the same number of neurons as the number of classes, for example for three class classifications, we use three neurons, for the classification of the four classes we use four neurons. Perceptron can use the activation function with option value 1 or -1 for output value from ANN.
Perceptron characteristics can be explained as follows.
Perceptron characteristics can be explained as follows.
Advertisement
- Perceptron can work well on data sets with linearly separable class distributions. That is why when faced with data sets whose class distribution is not linear, Perceptron can not do a good classification so for non-linear data set problems, ANN is designed to use multiple layers.
- Like ANN in general, Perceptron is capable of handling redundant data features because identical weights will be learned during the training process. The redundant feature weights will be pressed so that it becomes very small.
- Very sensitive to noise.
- The training process for model formation takes a long time.
- Result of model decision limit
No comments