A Rule-Based Approach to Neural Network Classifiers

Intelligent Systems in Accounting, Finance and Management

Smyth

1993

Self Cite

This paper describes our approach to the problem of automated knowledge acquisition from large databases of examples using an information-theoretic approach. Our previous research has resulted in practical algorithms (lTRULE) for the automatic induction of rules from large example databases. Utilizing these algorithms, the raw data can be transformed into a set of human readable IF THEN rules, thus giving insight into the knowledge hidden within the data. These rules can then be automatically loaded into an expert system shell. Alternatively, they can be used to build a new type of parallel inference system-a rule-based neural network. This process enables a prototype expert system to be automatically generated and up and running in a matter of minutes, compared with months using a manual knowledge-acquisition approach. The resulting expert system can then be used as a sophisticated search and analysis tool to query the original database capable of reasoning with uncertain and incomplete data.

Section: Rule-based Neural Networkmentioning

confidence: 99%

Automated Induction of Rule‐based Neural Networks from Databases

Intelligent Systems in Accounting, Finance and Management

Smyth

1993

Self Cite

“…Once the posterior probabilities are estimated, we need only choose the largest probability to make our classification decision. The above formula provides a simple method of constructing a neural network [8]. Consider the network of figure 1.…”

Section: A Neural Network For Classificationmentioning

confidence: 99%

Incremental learning with rule-based neural networks

Higgins

IJCNN-91-Seattle International Joint Conference on Neural Networks

A classifier for discrete-valued variable classification problems is presented. The system utilizes an information-theoretic algorithm for constructing informative rules from example data. These rules are then used to construct a neural network to perform parallel inference and posterior probability estimation. The network can be 'grown' incrementally, so that new data can be incorporated without repeating the training on previous data. It is shown that this technique performs comparably with other techniques such as back-propagation while having unique advantages in incremental learning capability, training efficiency, knowledge representation, and hardware implementation suitability.

“…Any classification scheme could be used. However, we utilize a rule -based information theoretic approach which is an extension of a first order Bayesian classifier, because of its ability to output probability estimates for the output classes [ 1 ]. The classifier defines correlations between input features and output classes as probabilistic rules of the form: If Y = y then X = x with prob.…”

Section: Supervised Learning Via a Rule-based Systemmentioning

confidence: 99%

“…In this work we suggest a general framework for classification, incorporating unsupervised and supervised learning, via a neural network (NN) model and a rule -based system [1,2] respectively. This is presented in Fig.…”

Section: Introductionmentioning

confidence: 99%

Texture analysis via unsupervised and supervised learning

Greenspan

IJCNN-91-Seattle International Joint Conference on Neural Networks

Chellappa

A framework for texture analysis is proposed based on combined unsupervised and supervised learning. The textured input is represented in the frequency-orientation space via a Gabor-wavelet pyramidal decomposition. This representation has optimal localization properties in space and in frequency, and is biologically motivated. In an unsupervised learning phase a neural network vector-quantization scheme is used for the quantization of the feature-vector attributes and a projection onto a reduced dimension clustered map for initial segmentation. A supervised stage follows in which labeling of the textured map is achieved using a rule based system. This information theoretic technique is utilized to find the most informative correlations between the attributes and the texture class specification. Ultimately, a minimal representation for a library of textures is learned in a training mode, following which the classification of new test images is achieved. This combined neural network-rule based approach is suggested as a general framework for pattern classification. Simulation results for the texture classification are given.