Prerequisits:

Second-cycle Bologna degree in the relevant track or a university (level VII) degree

Content (Syllabus outline):

1. Introduction to artificial intelligence
2. Data in geosciences and its specifics:
   • compositional and
   • orientational data,
   • outliers,
   • attributive data
3. Neural networks:
   • properties,
   • types,
   • topologies,
   • feed-forward and
   • recurrent networks
4. Neural network machine learning paradigms:
   • supervised,
   • unsupervised,
   • reinforced,
   • evolutionary
5. Data preparation and validation methods
6. Positive and negative aspects of neural networks
7. Approach to the problem
8. Examples from practice
9. Other methods of artificial intelligence:
   • linear and multiple regression,
   • decision trees,
   • support vector machines,
   • fuzzy logic
10. individual work

Readings:

• Nielsen A.M. Neural Networks and deep learning. Determination press, 2015, 224 p.
• https://static.latexstudio.net/article/2018/0912/neuralnetworksanddeeplearning.pdf
• http://neuralnetworksanddeeplearning.com/
• Haykin S. Neural Networks and learning machines, 3rd ed. Prentice Hall, 2009. https://dai.fmph.uniba.sk/courses/NN/haykin.neural-networks.3ed.2009.pdf
• ŽIBRET, Gorazd, ŠAJN, Robert. Hunting for geochemical associations of elements: factor analysis and self-organising maps. Mathematical geology. 2010, vol. 42, no. 6, str. 681-703. DOI: 10.1007/s11004-010-9288-3.
• ŽIBRET, Gorazd, ŠAJN, Robert, ALIJAGIĆ, Jasminka, STAFILOV, Trajče. Use of neural networks in the geochemical data interpretation. Zeitschrift für geologische Wissenschaften. 2012, bd. 40, h. 4/5, str. 253-266.
• CERAR, Sonja, MEZGA, Kim, ŽIBRET, Gorazd, URBANC, Janko, KOMAC, Marko. Comparison of prediction methods for oxygen-18 isotope composition in shallow groundwater. Science of the total environment. 2018, vol. 631-632, str. 358-368. DOI: 10.1016/j.scitotenv.2018.03.033.

Objectives and competences:

Student knows principles of artificial intelligence, and can use neural networks for solving problems

Intended learning outcomes:

The student is able to independently apply neural network methods using the MemBrain simulator. He/she knows how to analyse a problem, determine suitable network topologies, prepare data, train the network, properly evaluate the resulting model and use it in practice.

Learning and teaching methods:

• Lectures
• Seminar
• Independent work assignments
• Consultations
• e-Learning