This learning method is a simple model, calculates the response of each entry on Neural networks and that compares it with its target value. If the calculated response differs from the target value, the weights of the neural network are adapted according to a learning rule. Example: Single-layer perceptron, Multi-layer perceptron 64. 8.3.2. Uncontrolled (Unsupervised) Learning: Neural networks learn by defining the characteristics of their problems. Some of the advantages of Neural networks while classifying are as follows: • Neural networks are more robust due to their weights.
• Neural networks improve their performance by learning. This can continue after the training is applied. • The error rate is low after proper training is done and thus the degree of accuracy is high. • Neural networks are most steady especially in noisy environments 65.
- Thesis Statement
- Structure and Outline
- Voice and Grammar
8.4.Advantages of Artificial Neural Networks (ANN) • A neural network performs tasks that a linear program cannot even do.
• When one item of the neural network fails, the networks continue parallel to each other without any problem. • A neural network learns once and then does not need to be reprogrammed. • Can be applied to every application 66.
28 8.5.Disadvantages of Artificial Neural Networks (ANN) Although the use of artificial neural networks is high, these networks have some problems. In ANN models, there are main problems such as: • Relative importance of input parameters • The speed of slow convergence and over-fitting problems. 67 29 CHAPTER NINE CONCLUSION The aim of this study is to try to predict the prevalence of heart attacks in patients who are likely to have a heart attack, such as Epilepsy, Schizophrenia, Parkinson etc. In Figure 2, the EEG signal was first taken from the patients, then these signals were analyzed, and digital filters were applied to them. Because of these classification operations, the reference signal was generated.
Using artificial neural networks, new signals loaded into the database will determine whether patients will have a heart attack. After these EEG data sets are trained, new signals were loaded to test the accuracy and the accuracy of this neural network will be determined. If the accuracy of the digital filters used is too high, it is the filter that is may suitable for our EEG signal.
Lastly, with this neural network architecture, heart attacks will be tried to be predicted. The accuracy obtained after all these studies are completed, it was compared with the different studies in the literature.