You are here: Home » Study Plan » Subject

Sciences

Subject: MACHINE LEARNING AND DEEP LEARNING (A.A. 2020/2021)

master degree course in PHYSICS – FISICA

Course year 1
CFU 6
Teaching units Unit Machine learning and deep learning
Related or Additional Studies (lesson)
  • TAF: Supplementary compulsory subjects SSD: ING-INF/05 CFU: 6
Teachers: Simone CALDERARA
Exam type oral
Evaluation final vote
Teaching language English
Contents download pdf download

Teachers

Simone CALDERARA

Overview

The course will cover most of the techniques for pattern and data analysis. The main purpose will be offer a deep knowledge of the most important classification algorithms and model for automatically interpreting heterogeneous data. Learning techniques and methods will be covered as well in conjunction with basic supervised and unsupervised learning theory and methods.

Admission requirements

Basic knowledge of statistics and linear algebra

Course contents

MACHINE LEARNING

Classification theory, taxonomy and metrics:
-classification taxonomy
-learning assumptions
-generalization
-metrics

Introduction to Bayesian Probability:
-Bayesian Statistics
-Bayes Classifier
-Naïve assumption

Linear Models for Classification:
-LDA
-Logistic Regression

Margin methods and SVM

Ensemble Methods:
-Bagging
-Boosting
-Decision Trees and Random Forest

Unsupervised Learning and Dimensionality Reduction
-Hierarchical clustering
-Kmeans
-Spectral clustering
-PCA and LLE

6CFU

DEEP LEARNING

Neural Networks Introduction:
-Gradient descent
-Perceptron and MLE

Advanced NN:
-Convolutional network
-RNN and sequential data processing

Unsupervised NN
-Autoencoders
-Generative NN VAE
Generative NN GAN

Reinforcement Learning:
-Function approximation and RL
-Deep reinforcement Learning

3CFU

Teaching methods

Slide from the lecturer. Lab session coding all the presented techniques in Python and Pytorch.

Assessment methods

Oral exam with both thory and math derivations. 3 Questions: one purely about classification theory one about the classifier and its math derivation one about the comparison of different techniques in solving a real life problem

Learning outcomes

Knowledge and Understanding:
Know and understand the key techniques of pattern recognition and machine learning for data analysis of heterogeneous nature .
Applying knowledge and understanding:
Know how to apply the main classification algorithms of data, temporal sequences of information and complex patterns such as images and know how to apply the main techniques of machine learning is of type supervised that unsupervised .
Autonomy of Findings
independent judgment in analyzing and designing complex systems , assessing the impact of information technology solutions in the application context , both in relation to the technical aspects organizational aspects and proving to actively participate in decision-making in interdisciplinary contexts .
Communication skills:
describe a heterogeneous stakeholders in a clear and understandable information , ideas , problems and solutions as well as technical aspects ;
Learning skills:
-the capacity to recognize the need for independent learning throughout the lifespan , given the high rate of technological and methodological innovation in the area of Computer Engineering ;
- Ability to independently acquire new expertise from the literature

Readings

The course will use the following textbooks freely available on the web
Machine Learning:
[ISL]: An Introduction to Statistical Learning. James, Witten, Hastie and Tibshirani.
[ESL]: The Elements of Statistical Learning, Second Edition. Hastie, Tibshirani and
Friedman.
Deep Learning:
[DLB]: Deep Learning. Ian Goodfellow and Yoshua Bengio and Aaron Courville. Suggested further readings:
[RL]: Reinforcement Learning Sutton, Barto.
[PRML]:Pattern Recognition and Machine Learning. Bishop Christopher. Readings are intended to be completed after classes.