Neural and Evolutionary Learning

Objectives

Transferring to the students knowledges about the possible ways in which Evolutionary Computation can help improving the performance of other Machine Learning methods, with particular reference to Artificial Neural Networks.

General characterization

Code

200294

Credits

4.0

Responsible teacher

Leonardo Vanneschi

Hours

Weekly - Available soon

Total - Available soon

Teaching language

Portuguese. If there are Erasmus students, classes will be taught in English

Prerequisites

TBD.

Bibliography

Teaching method

Use of (black-)board and slides in the theoretical classes. Use of Python computer programming environments for the practical classes.

Evaluation method

TBD

Subject matter

  1. A quick revision of Evolutionary Algorithms
  2. A deeper look at Genetic Programming
  3. What can be evolved? The importance of representation
  4. Evolutionary Computation as a meta-learning algorithm
  5. A quick revision of the functioning of artificial neurons
  6. A look back at the Backpropagation
  7. From gradient descent to the Backpropagation: a step-by-step derivation of the algorithm
  8. Limitations of the Backpropagation
  9. How to Improve the Backpropagation?
  10. Neuro-evolution: Neural Networks evolved by means of Genetic Programming
  11. Comparison between Backpropagation and Neuro-evolution: pros and cons
  12. A hint on model’s interpretability and interpretable AI