Loading...
Please wait, while we are loading the content...
Similar Documents
Performance Issues of a Hybrid Symbolic, Connectionist Algorithm
| Content Provider | Scilit |
|---|---|
| Author | Hall, Lawrence O. Romaniuk, Steve G. |
| Copyright Year | 2020 |
| Description | This chapter considers the development of a hybrid symbolic connectionist learning system, called SC-net, which grows its own connectionist network structure based on the distinct examples presented to it. The system is applied to data from several different domains, ranging from the classical symbolic domain of soybean identification to the difficult non-linearly separable two-spiral problem. The latter problem is often a benchmark technique for neural network algorithms. The performance of the initial version of the learning system is very good in most non-symbolic domains and only acceptable in some of the symbolic domains. An analysis is done on the learning algorithm's complexity. A global attribute covering algorithm is added to the learning algorithm to increase performance in symbolic domains. This provides a more compact network representation and higher accuracy in some domains. Its usefulness and tradeoffs are shown. The rules generated from the system are presented and analyzed for semiconductor diagnosis-learning domain. Book Name: Hybrid Architectures for Intelligent Systems |
| Related Links | https://api.taylorfrancis.com/content/chapters/edit/download?identifierName=doi&identifierValue=10.1201/9781003068075-7&type=chapterpdf |
| Ending Page | 133 |
| Page Count | 29 |
| Starting Page | 105 |
| DOI | 10.1201/9781003068075-7 |
| Language | English |
| Publisher | Informa UK Limited |
| Publisher Date | 2020-09-09 |
| Access Restriction | Open |
| Subject Keyword | Book Name: Hybrid Architectures for Intelligent Systems Neural Network Algorithms Connectionist Structure Symbolic Domains Learning Algorithm |
| Content Type | Text |
| Resource Type | Chapter |
