Neurex is an expert system based on a multi-layered neural network. The era of neural networks and connectionism offers a new perspective on obtaining reliable knowledge for decision support and its user-friendly application. Traditional expert systems, which are rule-based and/or frame-based, often face challenges in creating a reliable knowledge base. Neural networks can overcome these difficulties. It's possible to create a knowledge base without experts, solely using data collections that describe the solved area, or with experts whose knowledge can be verified during the learning process. The expert system's usage process can be outlined as follows:

1. Definition of the Neural Network Topology: This step involves defining the number of input and output facts, as well as determining the number of hidden layers.
2. Formulation of Input and Output Facts (Attributes): Each fact is linked to a neuron in the input or output layer. The range of values for each attribute is also defined.
3. Definition of the Training Set: Patterns are entered using truth values (e.g., 0-100%) or values from the range defined in the previous steps.
4. Learning Phase of the Network: The weights of the connections (synapses) between neurons, the slopes of the sigmoid functions, and the thresholds of the neurons are computed using the Back Propagation (BP) method. Options are available to define parameters for this process, such as the learning rate and the number of learning cycles. These values form the expert system's memory or knowledge base. The learning process's results are displayed using the mean squared error, and the index of the worst pattern and its percentage error is also shown.
5. Consultation with the System: In this phase, the values of the input facts are defined, after which the values of the output facts are immediately deduced.