Artificial neural networks (ANNs) present one of the most important and widely used paradigms of artificial intelligence. Thus, this subject aims to enable PhD students for independent development, modelling and application of artificial neural networks in domain of complexity of intelligent machine systems through theoretical and practical aspects of machine learning algorithms and neural networks training.

The learning outcome of this subject considers the suitability for modelling and predicting changes of functional characteristics of systems and processes besides the introduction of PhD students into the basic methodology of complex problems modelling in mechanical engineering by the use of artificial neural networks which are able to learn and generalize the nature of phenomena on the basis of known experimental data. They can be trained in such way to find the solution, recognize the behaviour models with adequate accuracy, classify data and predict future events.

Theoretical education is organized into several parts: • Intelligent formalized methodologies; computational intelligence techniques - Adaptive processing and role of artificial neural networks (ANNs) in development of computational intelligence techniques, the history of development of ANNs; • ANNs-basic concepts - Structure of ANNs, processing element-neuron, activation function, learning algorithms of ANNs, simulation and processing of neural networks; • Models of ANNs - basic paradigms and examples; • Homogeneous ANNs - Perceptron, Back-propagation (BP) neural network, ART neural networks, Self-organizing map (SOM), etc.; • Heterogeneous ANNs (membrane potential, neural model, neural controller).

Practical education is organized into several parts: • ANNs in intelligent technologies - formalized conceptual design, group technology, feature recognition and part representation, advanced process planning, scheduling, recognition systems - image processing and analysis, monitoring and diagnostic of manufacturing processes, intelligent control of robots and machine systems, application in business and finances; • Developed software and their application - BPnet, ART-Simulator, Python 3.14.0rc2 and 3.13.7, MATLAB, Neuro Solutions, etc.

MSc degree of technically oriented faculty.

[1] Z. Miljković, D. Aleksendrić, ARTIFICIAL NEURAL NETWORKS – solved examples with theoretical background (2nd ed.), Textbook, University of Belgrade, Faculty of Mechanical Engineering, 2018, 18.1 /In Serbian/ [2] Z. Miljković, Systems of artificial neural networks in production technologies, Series IMS, Vol. 8, University of Belgrade, Faculty of Mechanical Engineering, 2003, 18.1 /In Serbian/ [3] Software packages (BPnet, ART-Simulator, Python 3.14.0rc2 and 3.13.7, MATLAB), Laboratory CeNT, University of Belgrade, Faculty of Mechanical Engineering, 18.13 [4] Laboratory mobile robot prototypes (K-Team's Khepera II mobile robot with gripper and camera; LEGO Mindstorms NXT and LEGO Mindstorms EV3 Sets of reconfigurable mobile robots equipped with sensors and micro-controllers), Laboratory CeNT, University of Belgrade, Faculty of Mechanical Engineering, 18.12 [5] Laboratory model of designed manufacturing system, Laboratory CeNT, University of Belgrade, Faculty of Mechanical Engineering, 18.12

Total assigned hours: 65

New material: 30
Elaboration and examples (recapitulation): 20

Auditory exercises: 0
Laboratory exercises: 0
Calculation tasks: 0
Seminar paper: 0
Project: 0
Consultations: 0
Discussion/workshop: 0
Research study work: 0

Review and grading of calculation tasks: 0
Review and grading of lab reports: 0
Review and grading of seminar papers: 10
Review and grading of the project: 0
Test: 0
Test: 0
Final exam: 5

Activity during lectures: 20
Test/test: 0
Laboratory practice: 0
Calculation tasks: 0
Seminar paper: 40
Project: 0
Final exam: 40
Requirement for taking the exam (required number of points): 40

Z. Miljković, D. Aleksendrić, (2018) ARTIFICIAL NEURAL NETWORKS–SOLVED EXAMPLES WITH THEORETICAL BACKGROUND (In Serbian), 2nd Ed., 225 pp. (ISBN 978-86-7083-961-8), UB–Faculty of Mech. Eng., Belgrade.; Freeman, J.A., Skapura, D.M., (1991) NEURAL NETWORKS – ALGORITHMS, APPLICATIONS AND PROGRAMMING TECHNIQUES, 401 pp. (ISBN 9780201513769), Addison-Wesley Publishing Company.; Golden, R.M., (1996) MATHEMATICAL METHODS FOR NEURAL NETWORK ANALYSIS AND DESIGN, 419 pp. (ISBN 9780262071741), Cambridge, Mass.: The MIT Press.; Skapura, D.M., (1996) BUILDING NEURAL NETWORKS, ACM Press Series, 286 pp. (ISBN 9780201539219), Addison-Wesley Professional.; Zalzala, A.M.S., Morris, A.S., (1996) NEURAL NETWORKS FOR ROBOTIC CONTROL -THEORY AND APPLICATIONS, VIII+278 pp. (ISBN 0131198920), London, Ellis Horwood Limited.