Unit NEURAL NETWORKS AND LEARNING ALGORITHMS
- Course
- Electronic engineering
- Study-unit Code
- A005870
- Curriculum
- Industrial electronics
- Teacher
- Renzo Perfetti
- Teachers
-
- Renzo Perfetti
- Hours
- 72 ore - Renzo Perfetti
- CFU
- 9
- Course Regulation
- Coorte 2025
- Offered
- 2025/26
- Learning activities
- Affine/integrativa
- Area
- Attività formative affini o integrative
- Academic discipline
- ING-IND/31
- Type of study-unit
- Opzionale (Optional)
- Type of learning activities
- Attività formativa monodisciplinare
- Language of instruction
- Italian
- Contents
- Introduction to neural networks and learning algorithms for classification and regression.
- Reference texts
- Duda, Hart , Stork, Pattern Classification, 2nd ed., Wiley C.M. Bishop, Pattern Recognition and Machine Learning, Springer James, Witten, Hastie, Tibshirani, An Introduction to Statistical Learning, Springer S. Haykin, Neural Networks, Prentice-Hall N. Cristianini, J. Shawe-Taylor, An Introduction to Support Vector Machines, Cambridge University Press
- Educational objectives
- Understanding of basic learning algorithms and of problems arising in their application and evaluation in pattern recognition and data analysis
- Prerequisites
- Probability and random variables. Linear algebra.
- Teaching methods
- Face to face lessons
- Learning verification modality
- Oral exam
- Extended program
- Review on quadratic forms and random variables. Basic facts on machine learning. Bayesian classification. Parametric and nonparametric methods. ML estimation. Gaussian mixtures. Expectation maximization. Linear separation. Perceptron. Steepest descent. LMS algorithm. Multiclass linear separation. Classification performance: overfitting, bias and variance, ROC, confusion matrix. Test and validation. Multilayer neural networks. Error backpropagation learning algorithm. Regolarization techniques. Hints on recurrent networks and deep learning. Support Vector Machines. Clustering algorithms. Self-organizing neural networks. Linear regression. Neural nets for regression.