Alessia SAGGESE | ARTIFICIAL INTELLIGENCE

cod. 0623300012

ARTIFICIAL INTELLIGENCE

	0623300012
	DEPARTMENT OF INFORMATION AND ELECTRICAL ENGINEERING AND APPLIED MATHEMATICS
	EQF7
	ELECTRICAL ENGINEERING FOR DIGITAL ENERGY
	2024/2025

PERCORSO COMUNE Cohort 2023/2024

	OBBLIGATORIO
	YEAR OF COURSE 2
	YEAR OF DIDACTIC SYSTEM 2023
	FULL ACADEMIC YEAR

TEACHING UNITS

SSD	CFU	HOURS	ACTIVITY	TYPE OF ACTIVITY
ING-INF/05	6	48	LESSONS	COMPULSORY SUBJECTS, CHARACTERISTIC OF THE CLASS
ING-INF/05	2	16	EXERCISES	COMPULSORY SUBJECTS, CHARACTERISTIC OF THE CLASS
ING-INF/05	4	32	LAB	COMPULSORY SUBJECTS, CHARACTERISTIC OF THE CLASS

PROFESSORS

	VINCENZO MATTA T Curriculum
	ALESSIA SAGGESE Curriculum
	ANTONIO DELLA CIOPPA Curriculum

	Objectives
	THE COURSE HAS THE TWOFOLD PURPOSE OF: I) ILLUSTRATING THE MAIN METHODOLOGIES OF INTEREST FOR STATISTICAL DATA ANALYSIS; II) APPLYING SUCH METHODOLOGIES TO RELEVANT PRACTICAL PROBLEMS, USING TOOLS COMMONLY EMPLOYED FOR STATISTICAL ANALYSIS, DATA VISUALIZATION AND PROCESSING. KNOWLEDGE AND UNDERSTANDING. •ACQUISITION OF THE MAIN STATISTICAL INFERENCE AND DATA ANALYSIS. •PARAMETRIC VS. NON PARAMETRIC APPROACHES. SUPERVISED VS. UNSUPERVISED APPROACHES. •ACQUISITION OF THE MAIN TECHNIQUES AND TOOLS FOR BIG DATA ANALYSIS. APPLYING KNOWLEDGE AND UNDERSTANDING. •ABILITY TO APPLY THE MAIN TECHNIQUES FOR STATISTICAL INFERENCE AND DATA ANALYSIS TO PRACTICAL PROBLEMS (E.G., SOCIAL OR BIOMEDICAL DATA). •ABILITY TO EXAMINE BIG DATA, ARRANGED IN RATHER COMPLEX AND/OR HETEROGENEOUS STRUCTURES • ABILITY TO USE SOFTWARE (E.G., R, PYTHON, MATLAB) FOR STATISTICAL DATA ANALYSIS, DATA VISUALIZATION AND PROCESSING. PART 2: LEARNING OBJECTIVES THE AIM OF MODULE 2 OF THE COURSE IS TO INTRODUCE STUDENTS TO THE BASIC PROBLEMS, MODELS AND TECHNIQUES OF ARTIFICIAL INTELLIGENCE (IA) AND TO ENABLE THEM TO MODEL AND IMPLEMENT AI APPLICATIONS FOR SOLVING SPECIFIC PROBLEMS IN THE DIGITAL ENERGY SECTOR. THE MODULE 2 OF COURSE COVERS THE FUNDAMENTAL CONCEPTS, MODELLING APPROACHES AND RESOLUTION METHODS OF CLASSICAL IA FOR THE IMPLEMENTATION OF ARTIFICIAL AGENTS CAPABLE OF SOLVING LEARNING, OPTIMISATION AND PREDICTION PROBLEMS, WITH PARTICULAR REFERENCE TO THE DIGITAL ENERGY SECTOR. MODULE 2: LEARNING OBJECTIVES THE AIM OF MODULE 2 OF THE COURSE IS TO INTRODUCE STUDENTS TO THE BASIC PROBLEMS, MODELS AND TECHNIQUES OF ARTIFICIAL INTELLIGENCE (IA) AND TO ENABLE THEM TO MODEL AND IMPLEMENT AI APPLICATIONS FOR SOLVING SPECIFIC PROBLEMS IN THE DIGITAL ENERGY SECTOR. THE MODULE 2 OF COURSE COVERS THE FUNDAMENTAL CONCEPTS, MODELLING APPROACHES AND RESOLUTION METHODS OF CLASSICAL IA FOR THE IMPLEMENTATION OF ARTIFICIAL AGENTS CAPABLE OF SOLVING LEARNING, OPTIMISATION AND PREDICTION PROBLEMS, WITH PARTICULAR REFERENCE TO THE DIGITAL ENERGY SECTOR. MODULE 2: KNOWLEDGE AND UNDERSTANDING SKILLS METHODOLOGIES AND TOOLS FOR MODELLING INTELLIGENT AGENTS. PARADIGM OF PROBLEM SOLVING PROBLEMS AS A SEARCH IN A SPACE OF STATES AND SOLUTIONS. LOCAL SEARCH. MACHINE LEARNING: SUPERVISED, UNSUPERVISED METHODS. NEURAL NETWORKS. EVOLUTIONARY COMPUTATION. MODULE 2: APPLIED KNOWLEDGE AND UNDERSTANDING IDENTIFY THE MOST SUITABLE MODELS AND TOOLS FOR THE REPRESENTATION AND SOLUTION OF PROBLEMS COMPLEX PROBLEMS ACCORDING TO DIFFERENT IA APPROACHES AND ESTIMATE THEIR COMPUTATIONAL COSTS AND PERFORMANCE. FORMULATE THE SOLUTION OF A PROBLEM IN TERMS OF A SEARCH IN A SPECIALLY DEFINED STATE SPACE. DESIGN INTELLIGENT AGENTS USING APPROPRIATE PROBLEM-SOLVING TECHNIQUES AS INSTANCES OF GENERAL CLASSES, WITH PARTICULAR REFERENCE TO THE DIGITAL ENERGY SECTOR.

THE COURSE HAS THE TWOFOLD PURPOSE OF: I) ILLUSTRATING THE MAIN METHODOLOGIES OF INTEREST FOR STATISTICAL DATA ANALYSIS; II) APPLYING SUCH METHODOLOGIES TO RELEVANT PRACTICAL PROBLEMS, USING TOOLS COMMONLY EMPLOYED FOR STATISTICAL ANALYSIS, DATA VISUALIZATION AND PROCESSING.

KNOWLEDGE AND UNDERSTANDING.
•ACQUISITION OF THE MAIN STATISTICAL INFERENCE AND DATA ANALYSIS.
•PARAMETRIC VS. NON PARAMETRIC APPROACHES. SUPERVISED VS. UNSUPERVISED APPROACHES.
•ACQUISITION OF THE MAIN TECHNIQUES AND TOOLS FOR BIG DATA ANALYSIS.

APPLYING KNOWLEDGE AND UNDERSTANDING.
•ABILITY TO APPLY THE MAIN TECHNIQUES FOR STATISTICAL INFERENCE AND DATA ANALYSIS TO PRACTICAL PROBLEMS (E.G., SOCIAL OR BIOMEDICAL DATA).
•ABILITY TO EXAMINE BIG DATA, ARRANGED IN RATHER COMPLEX AND/OR HETEROGENEOUS STRUCTURES
• ABILITY TO USE SOFTWARE (E.G., R, PYTHON, MATLAB) FOR STATISTICAL DATA ANALYSIS, DATA VISUALIZATION AND PROCESSING.

PART 2: LEARNING OBJECTIVES

THE AIM OF MODULE 2 OF THE COURSE IS TO INTRODUCE STUDENTS TO THE BASIC PROBLEMS, MODELS AND TECHNIQUES OF ARTIFICIAL INTELLIGENCE (IA) AND TO ENABLE THEM TO MODEL AND IMPLEMENT AI APPLICATIONS FOR SOLVING SPECIFIC PROBLEMS IN THE DIGITAL ENERGY SECTOR.
THE MODULE 2 OF COURSE COVERS THE FUNDAMENTAL CONCEPTS, MODELLING APPROACHES AND RESOLUTION METHODS
OF CLASSICAL IA FOR THE IMPLEMENTATION OF ARTIFICIAL AGENTS CAPABLE OF SOLVING LEARNING, OPTIMISATION AND PREDICTION PROBLEMS, WITH PARTICULAR REFERENCE TO THE DIGITAL ENERGY SECTOR.

MODULE 2: LEARNING OBJECTIVES

THE AIM OF MODULE 2 OF THE COURSE IS TO INTRODUCE STUDENTS TO THE BASIC PROBLEMS, MODELS AND TECHNIQUES OF ARTIFICIAL INTELLIGENCE (IA) AND TO ENABLE THEM TO MODEL AND IMPLEMENT AI APPLICATIONS FOR SOLVING SPECIFIC PROBLEMS IN THE DIGITAL ENERGY SECTOR.
THE MODULE 2 OF COURSE COVERS THE FUNDAMENTAL CONCEPTS, MODELLING APPROACHES AND RESOLUTION METHODS
OF CLASSICAL IA FOR THE IMPLEMENTATION OF ARTIFICIAL AGENTS CAPABLE OF SOLVING LEARNING, OPTIMISATION AND PREDICTION PROBLEMS, WITH PARTICULAR REFERENCE TO THE DIGITAL ENERGY SECTOR.

MODULE 2: KNOWLEDGE AND UNDERSTANDING SKILLS

METHODOLOGIES AND TOOLS FOR MODELLING INTELLIGENT AGENTS. PARADIGM OF PROBLEM SOLVING
PROBLEMS AS A SEARCH IN A SPACE OF STATES AND SOLUTIONS. LOCAL SEARCH.
MACHINE LEARNING: SUPERVISED, UNSUPERVISED METHODS. NEURAL NETWORKS. EVOLUTIONARY COMPUTATION.

MODULE 2: APPLIED KNOWLEDGE AND UNDERSTANDING

IDENTIFY THE MOST SUITABLE MODELS AND TOOLS FOR THE REPRESENTATION AND SOLUTION OF PROBLEMS
COMPLEX PROBLEMS ACCORDING TO DIFFERENT IA APPROACHES AND ESTIMATE THEIR COMPUTATIONAL COSTS AND PERFORMANCE.
FORMULATE THE SOLUTION OF A PROBLEM IN TERMS OF A SEARCH IN A SPECIALLY DEFINED STATE SPACE.
DESIGN INTELLIGENT AGENTS USING APPROPRIATE PROBLEM-SOLVING TECHNIQUES AS INSTANCES OF GENERAL CLASSES, WITH PARTICULAR REFERENCE TO THE DIGITAL ENERGY SECTOR.

	Prerequisites
	PREREQUISITES: SUITABLE KNOWLEDGE OF MATHEMATICS AND FUNDAMENTALS OF PROBABILITY AND STATISTICS. MODULE 2 PREREQUISITES: THE COURSE REQUIRES KNOWLEDGE OF THE PYTHON PROGRAMMING LANGUAGE.

	Contents
	DIDACTIC UNIT 1: INTRODUCTION AND PARAMETRIC METHODS (LECTURE/PRACTICE/LABORATORY HOURS 6/0/2) - 1 (2 HOURS LECTURE): INTRODUCTION TO DATA ANALYSIS. PREDICTION VS. INFERENCE. REGRESSION VS. CLASSIFICATION. PARAMETRIC METHODS AND MAXIMUM LIKELIHOOD (ML) ESTIMATION. - 2 (2 HOURS LECTURE): BAYESIAN APPROACH AND MINIMUM-MEAN-SQUARED-ERROR (MMSE). COST FUNCTIONS FOR ESTIMATION AND REGRESSION PROBLEMS. - 3 (2 HOURS LECTURE): ML AND MMSE ESTIMATORS FOR CLASSIC GAUSSIAN PROBLEMS. - 4 (2 HOURS LABORATORY): COMPUTER-AIDED IMPLEMENTATION AND PERFORMANCE EVALUATION OF THE ESTIMATORS DISCUSSED IN THE PREVIOUS LECTURES. KNOWLEDGE AND UNDERSTANDING: PARAMETRIC ESTIMATION METHODS FOR STATISTICAL LEARNING PROBLEMS. APPLYING KNOWLEDGE AND UNDERSTANDING: DESIGNING AND IMPLEMENTING PARAMETRIC ESTIMATION ALGORITHMS. DIDACTIC UNIT 2: SUPERVISED LEARNING METHODS FOR REGRESSION (LECTURE/PRACTICE/LABORATORY HOURS 16/0/6) - 5 (2 HOURS LECTURE): REGRESSION FUNCTION AND SUPERVISED PARAMETRIC MODELS. - 6 (2 HOURS LECTURE): LINEAR REGRESSION. - 7 (2 HOURS LECTURE): STATISTICAL INFERENCE. HYPOTHESIS TESTS AND P-VALUE. - 8 (2 HOURS LECTURE): VARIABLE SELECTION. STEPWISE PROCEDURES. - 9 (2 HOURS LECTURE): DATA NORMALIZATION. REGULARIZATION/SHRINKAGE STRATEGIES. MULTICOLLINEARITY AND HIGH DIMENSIONALITY. RIDGE REGRESSION. - 10 (2 HOURS LECTURE): LASSO METHOD. DIMENSIONALITY REDUCTION. - 11 (2 HOURS LABORATORY): COMPUTER-AIDED IMPLEMENTATION OF SIMPLE AND MULTIPLE LINEAR REGRESSION ALGORITHMS. - 12 (2 HOURS LABORATORY): COMPUTER-AIDED IMPLEMENTATION OF INFERENTIAL STRATEGIES, RIDGE AND LASSO. - 13 (2 HOURS LECTURE): CROSS-VALIDATION. - 14 (2 HOURS LECTURE): SUPERVISED NON-PARAMETRIC METHODS. LOCAL METHODS. NAÏVE-KERNEL. K-NN METHOD. - 15 (2 HOURS LABORATORY): COMPUTER-AIDED IMPLEMENTATION OF NAÏVE-KERNEL AND K-NN METHODS. KNOWLEDGE AND UNDERSTANDING: REGRESSION MODELS. ESTIMATION OF MODEL PARAMETERS, VARIABLE SELECTION, AND SIGNIFICANCE TESTS TO DETERMINE THE INFLUENCE FACTORS AND PERFORM MODEL INTERPRETABILITY. REGULARIZATION TECHNIQUES IN HIGH-DIMENSIONAL PROBLEMS. APPLYING KNOWLEDGE AND UNDERSTANDING: DESIGNING AND IMPLEMENTING REGRESSION AND INFERENCE ALGORITHMS, FOR PREDICTION, DATA INTERPRETATION AND EVALUATION OF THE STATISTICAL SIGNIFICANCE OF THE RESULTS. DIDACTIC UNIT 3: CLASSIFICATION (LECTURE/PRACTICE/LABORATORY HOURS 10/0/8) - 16 (2 HOURS LECTURE): PARAMETRIC DECISION METHODS. NEYMAN-PEARSON CRITERION AND BAYESIAN APPROACH. - 17 (2 HOURS LECTURE): PARAMETRIC DETECTORS FOR A CLASSIC GAUSSIAN PROBLEM. - 18 (2 HOURS LABORATORY): COMPUTER-AIDED IMPLEMENTATION OF THE PARAMETRIC DETECTORS FOR THE GAUSSIAN PROBLEM ILLUSTRATED IN THE PREVIOUS LECTURE. - 19 (2 HOURS LECTURE): SUPERVISED METHODS. NAÏVE-BAYES AND LOGISTIC REGRESSION. - 20 (2 HOURS LECTURE): GRADIENT DESCENT ALGORITHMS FOR REGRESSION AND CLASSIFICATION PROBLEMS. - 21 (2 HOURS LABORATORY): COMPUTER-AIDED IMPLEMENTATION OF GRADIENT DESCENT ALGORITHMS FOR LOGISTIC REGRESSION. - 22 (2 HOURS LECTURE): LINEAR DISCRIMINANT ANALYSIS (LDA). - 23 (2 HOURS LABORATORY): COMPUTER-AIDED IMPLEMENTATION OF THE CLASSIFICATION METHODS ILLUSTRATED DURING THE COURSE. - 24 (2 HOURS LABORATORY): COMPUTER-AIDED IMPLEMENTATION OF THE CLASSIFICATION METHODS ILLUSTRATED DURING THE COURSE. KNOWLEDGE AND UNDERSTANDING: STRATEGIES FOR CLASSIFICATION PROBLEMS. OPTIMIZATION ALGORITHMS FOR STATISTICAL LEARNING (E.G., GRADIENT-DESCENT AND STOCHASTIC-GRADIENT-DESCENT ALGORITHMS). APPLYING KNOWLEDGE AND UNDERSTANDING: DESIGNING AND IMPLEMENTING CLASSIFICATION ALGORITHMS. IMPLEMENTING DISTRIBUTED DATA-ANALYSIS ALGORITHMS BY MEANS OF SUITABLE FRAMEWORKS. TOTAL LECTURE/PRACTICE/LABORATORY HOURS 32/0/16 MODULE 2 - TEACHING UNIT 1: INTRODUCTION TO ARTIFICIAL INTELLIGENCE (LECTURE/PRACTICE/LABORATORY HOURS 2/0/0) - 1 (2 HOURS LECTURE): BASIC CONCEPTS: AGENT, ENVIRONMENT, AUTONOMY, EVALUATION AND FEEDBACK, LEARNING. PEAS MODEL KNOWLEDGE AND UNDERSTANDING: UNDERSTANDING OF THE ARTIFICIAL AGENT CONCEPT AND THE PEAS MODEL APPLIED KNOWLEDGE AND UNDERSTANDING: KNOWING HOW TO DEFINE THE PEAS MODEL OF INTELLIGENT AGENTS FOR SPECIFIC APPLICATIONS MODULE 2 - LEARNING UNIT 2: PROBLEM SOLVING WITH RESEARCH (HRS. LECTURE/PRACTICE/LABORATORY 4/0/2) - 2 (2 HOURS LECTURE): PROBLEM SOLVING: INFORMED SEARCH STRATEGIES AND HEURISTIC FUNCTIONS - 3 (2 HOURS LECTURE): LOCAL SEARCH AND OPTIMISATION PROBLEMS: HILL CLIMBER AND SIMULATED ANNEALING - 4 (2 HOURS LABORATORY): USE OF LOCAL SEARCH TO SOLVE OPTIMISATION PROBLEMS KNOWLEDGE AND UNDERSTANDING: KNOWLEDGE OF THE VARIOUS SEARCH STRATEGIES FOR SOLVING SEARCH AND OPTIMISATION PROBLEMS APPLIED KNOWLEDGE AND UNDERSTANDING: KNOWING HOW TO CHOOSE THE MOST SUITABLE STRATEGY FOR SOLVING SEARCH AND OPTIMISATION PROBLEMS, WITH PARTICULAR REFERENCE TO THE DIGITAL ENERGY SECTOR. MODULE 2 - TEACHING UNIT 3: MACHINE LEARNING (LECTURE/PRACTICE/LABORATORY HOURS 6/0/4) - 5 (2 HOURS LECTURE): SUPERVISED AND UNSUPERVISED MACHINE LEARNING. REGRESSION, CLASSIFICATION, PREDICTION - 6 (2 HOURS LECTURE): LINEAR REGRESSION AND LOGISTIC REGRESSION - 7 (2 HOURS LECTURE): DECISION TREES AND RANDOM FORESTS - 8 (2 HOURS LABORATORY): APPLICATION OF MACHINE LEARNING TECHNIQUES TO REGRESSION, CLASSIFICATION AND PREDICTION PROBLEMS - 9 (2 HOURS LABORATORY): APPLICATION OF MACHINE LEARNING TECHNIQUES TO REGRESSION, CLASSIFICATION AND PREDICTION PROBLEMS, WITH PARTICULAR REFERENCE TO THE DIGITAL ENERGY SECTOR. KNOWLEDGE AND APPLIED COMPREHENSION: KNOWLEDGE OF VARIOUS LEARNING STRATEGIES FOR SOLVING REGRESSION, CLASSIFICATION AND PREDICTION PROBLEMS, WITH PARTICULAR REFERENCE TO THE DIGITAL ENERGY SECTOR. APPLIED KNOWLEDGE AND UNDERSTANDING: KNOWING HOW TO CHOOSE THE MOST SUITABLE STRATEGY FOR SOLVING LEARNING PROBLEMS, WITH PARTICULAR REFERENCE TO THE DIGITAL ENERGY SECTOR. MODULE 2 - TEACHING UNIT 4: EVOLUTIONARY COMPUTATION (LECTURE/PRACTICE/LABORATORY HOURS 6/4/4) - 10 (2 HOURS LECTURE): FUNDAMENTALS OF EVOLUTIONARY BIOLOGY: SELECTION, RECOMBINATION AND MUTATION - INTRODUCTION TO EVOLUTIONARY ALGORITHMS - 11 (2 HOURS LECTURE): GENETIC ALGORITHMS AND DIFFERENTIAL EVOLUTION - 12 (2 HOURS LECTURE): GENETIC PROGRAMMING - 13 (2 HOURS LABORATORY): APPLICATION OF GENETIC ALGORITHMS AND DIFFERENTIAL EVOLUTION TO OPTIMISATION PROBLEMS - 14 (2 HOURS LABORATORY): APPLICATION OF GENETIC PROGRAMMING TO OPTIMISATION AND LEARNING PROBLEMS - 15 (2 HOURS PRACTICE): DISCUSSION OF PROJECT WORK. - 16 (2 HOURS PRACTICE): DISCUSSION OF PROJECT WORK. KNOWLEDGE AND ABILITY TO UNDERSTAND: TO ACQUIRE KNOWLEDGE OF THE FUNDAMENTAL ELEMENTS OF EVOLUTIONARY COMPUTATION APPLIED KNOWLEDGE AND UNDERSTANDING: TO KNOW HOW TO APPLY EVOLUTIONARY COMPUTATION TO VARIOUS TYPES OF REAL-WORLD PROBLEMS, WITH PARTICULAR REFERENCE TO THE DIGITAL ENERGY SECTOR. TO KNOW HOW TO USE THE PYTHON FRAMEWORKS FOR THE REALISATION OF EVOLUTIONARY ALGORITHMS, AND TO CHOOSE THE PARAMETERS OF SUCH NETWORKS APPROPRIATELY. MODULE 2 - TEACHING UNIT 5: NEURAL NETWORKS (LECTURE/PRACTICE/LABORATORY HOURS 8/2/6) - 17 (2 LECTURE HOURS): INTRODUCTION TO NEURAL NETWORKS: THE MCCULLOCH AND PITTS NEURON. THE ROSENBLATT PERCEPTRON. FEED-FORWARD ARCHITECTURES. FULLY-CONNECTED AND SPARSE-CONNECTED NETWORKS. MULTI-LAYER PERCEPTRON. - 18 (2 HOURS LECTURE): FEEDFORWARD ARCHITECTURE, ACTIVATION FUNCTIONS AND THEIR CHARACTERISTICS, BACKPROPAGATION AND THE STOCHASTIC GRADIENT DESCENT ALGORITHM. - 19 (2 HOURS LECTURE): EARLY STOPPING, MOMENTUM, ADAPTIVE LEARNING RATE, REGULARISATION. - 20 (2 HOURS LECTURE): COMPETITIVE NEURAL NETWORKS. LEARNING VECTOR QUANTIZATION NETWORKS. - 21 (2 HOURS LABORATORY): USE OF MLP NETWORKS IN CLASSIFICATION PROBLEMS. - 22 (2 HOURS LABORATORY): USE OF MLP NETWORKS IN REGRESSION PROBLEMS. - 23 (2 HOURS LABORATORY): USE OF LVQ NETWORKS IN CLASSIFICATION PROBLEMS. - 24 (2 HOURS PRACTICE): DISCUSSION PROGRESS PROJECT WORK. KNOWLEDGE AND UNDERSTANDING: THE NEURAL LEARNING PARADIGM. ARCHITECTURE AND FUNCTIONING OF MLP AND LVQ NETWORKS. APPLIED KNOWLEDGE AND UNDERSTANDING: KNOWING HOW TO DESIGN MACHINE LEARNING SOLUTIONS BASED ON THE NEURAL PARADIGM, WITH PARTICULAR REFERENCE TO THE DIGITAL ENERGY SECTOR. KNOWING HOW TO USE THE PYTHON FRAMEWORK FOR THE REALISATION OF NEURAL NETWORKS, AND HOW TO CHOOSE THE PARAMETERS OF SUCH NETWORKS APPROPRIATELY. MODULE 2 TOTAL HOURS LECTURE/PRACTICE/LABORATORY 26/6/16

Contents

DIDACTIC UNIT 1: INTRODUCTION AND PARAMETRIC METHODS
(LECTURE/PRACTICE/LABORATORY HOURS 6/0/2)

- 1 (2 HOURS LECTURE): INTRODUCTION TO DATA ANALYSIS. PREDICTION VS. INFERENCE. REGRESSION VS. CLASSIFICATION. PARAMETRIC METHODS AND MAXIMUM LIKELIHOOD (ML) ESTIMATION.
- 2 (2 HOURS LECTURE): BAYESIAN APPROACH AND MINIMUM-MEAN-SQUARED-ERROR (MMSE). COST FUNCTIONS FOR ESTIMATION AND REGRESSION PROBLEMS.
- 3 (2 HOURS LECTURE): ML AND MMSE ESTIMATORS FOR CLASSIC GAUSSIAN PROBLEMS.
- 4 (2 HOURS LABORATORY): COMPUTER-AIDED IMPLEMENTATION AND PERFORMANCE EVALUATION OF THE ESTIMATORS DISCUSSED IN THE PREVIOUS LECTURES.

KNOWLEDGE AND UNDERSTANDING:
PARAMETRIC ESTIMATION METHODS FOR STATISTICAL LEARNING PROBLEMS.
APPLYING KNOWLEDGE AND UNDERSTANDING:
DESIGNING AND IMPLEMENTING PARAMETRIC ESTIMATION ALGORITHMS.

DIDACTIC UNIT 2: SUPERVISED LEARNING METHODS FOR REGRESSION (LECTURE/PRACTICE/LABORATORY HOURS 16/0/6)

- 5 (2 HOURS LECTURE): REGRESSION FUNCTION AND SUPERVISED PARAMETRIC MODELS.
- 6 (2 HOURS LECTURE): LINEAR REGRESSION.
- 7 (2 HOURS LECTURE): STATISTICAL INFERENCE. HYPOTHESIS TESTS AND P-VALUE.
- 8 (2 HOURS LECTURE): VARIABLE SELECTION. STEPWISE PROCEDURES.
- 9 (2 HOURS LECTURE): DATA NORMALIZATION. REGULARIZATION/SHRINKAGE STRATEGIES. MULTICOLLINEARITY AND HIGH DIMENSIONALITY. RIDGE REGRESSION.
- 10 (2 HOURS LECTURE): LASSO METHOD. DIMENSIONALITY REDUCTION.
- 11 (2 HOURS LABORATORY): COMPUTER-AIDED IMPLEMENTATION OF SIMPLE AND MULTIPLE LINEAR REGRESSION ALGORITHMS.
- 12 (2 HOURS LABORATORY): COMPUTER-AIDED IMPLEMENTATION OF INFERENTIAL STRATEGIES, RIDGE AND LASSO.
- 13 (2 HOURS LECTURE): CROSS-VALIDATION.
- 14 (2 HOURS LECTURE): SUPERVISED NON-PARAMETRIC METHODS. LOCAL METHODS. NAÏVE-KERNEL. K-NN METHOD.
- 15 (2 HOURS LABORATORY): COMPUTER-AIDED IMPLEMENTATION OF NAÏVE-KERNEL AND K-NN METHODS.

KNOWLEDGE AND UNDERSTANDING:
REGRESSION MODELS. ESTIMATION OF MODEL PARAMETERS, VARIABLE SELECTION, AND SIGNIFICANCE TESTS TO DETERMINE THE INFLUENCE FACTORS AND PERFORM MODEL INTERPRETABILITY. REGULARIZATION TECHNIQUES IN HIGH-DIMENSIONAL PROBLEMS.
APPLYING KNOWLEDGE AND UNDERSTANDING:
DESIGNING AND IMPLEMENTING REGRESSION AND INFERENCE ALGORITHMS, FOR PREDICTION, DATA INTERPRETATION AND EVALUATION OF THE STATISTICAL SIGNIFICANCE OF THE RESULTS.

DIDACTIC UNIT 3: CLASSIFICATION (LECTURE/PRACTICE/LABORATORY HOURS 10/0/8)

- 16 (2 HOURS LECTURE): PARAMETRIC DECISION METHODS. NEYMAN-PEARSON CRITERION AND BAYESIAN APPROACH.
- 17 (2 HOURS LECTURE): PARAMETRIC DETECTORS FOR A CLASSIC GAUSSIAN PROBLEM.
- 18 (2 HOURS LABORATORY): COMPUTER-AIDED IMPLEMENTATION OF THE PARAMETRIC DETECTORS FOR THE GAUSSIAN PROBLEM ILLUSTRATED IN THE PREVIOUS LECTURE.
- 19 (2 HOURS LECTURE): SUPERVISED METHODS. NAÏVE-BAYES AND LOGISTIC REGRESSION.
- 20 (2 HOURS LECTURE): GRADIENT DESCENT ALGORITHMS FOR REGRESSION AND CLASSIFICATION PROBLEMS.
- 21 (2 HOURS LABORATORY): COMPUTER-AIDED IMPLEMENTATION OF GRADIENT DESCENT ALGORITHMS FOR LOGISTIC REGRESSION.
- 22 (2 HOURS LECTURE): LINEAR DISCRIMINANT ANALYSIS (LDA).
- 23 (2 HOURS LABORATORY): COMPUTER-AIDED IMPLEMENTATION OF THE CLASSIFICATION METHODS ILLUSTRATED DURING THE COURSE.
- 24 (2 HOURS LABORATORY): COMPUTER-AIDED IMPLEMENTATION OF THE CLASSIFICATION METHODS ILLUSTRATED DURING THE COURSE.

KNOWLEDGE AND UNDERSTANDING:
STRATEGIES FOR CLASSIFICATION PROBLEMS. OPTIMIZATION ALGORITHMS FOR STATISTICAL LEARNING (E.G., GRADIENT-DESCENT AND STOCHASTIC-GRADIENT-DESCENT ALGORITHMS).
APPLYING KNOWLEDGE AND UNDERSTANDING:
DESIGNING AND IMPLEMENTING CLASSIFICATION ALGORITHMS. IMPLEMENTING DISTRIBUTED DATA-ANALYSIS ALGORITHMS BY MEANS OF SUITABLE FRAMEWORKS.

TOTAL LECTURE/PRACTICE/LABORATORY HOURS 32/0/16

MODULE 2 - TEACHING UNIT 1: INTRODUCTION TO ARTIFICIAL INTELLIGENCE
(LECTURE/PRACTICE/LABORATORY HOURS 2/0/0)
- 1 (2 HOURS LECTURE): BASIC CONCEPTS: AGENT, ENVIRONMENT, AUTONOMY, EVALUATION AND FEEDBACK, LEARNING. PEAS MODEL
KNOWLEDGE AND UNDERSTANDING: UNDERSTANDING OF THE ARTIFICIAL AGENT CONCEPT AND THE PEAS MODEL
APPLIED KNOWLEDGE AND UNDERSTANDING: KNOWING HOW TO DEFINE THE PEAS MODEL OF INTELLIGENT AGENTS FOR SPECIFIC APPLICATIONS

MODULE 2 - LEARNING UNIT 2: PROBLEM SOLVING WITH RESEARCH
(HRS. LECTURE/PRACTICE/LABORATORY 4/0/2)
- 2 (2 HOURS LECTURE): PROBLEM SOLVING: INFORMED SEARCH STRATEGIES AND HEURISTIC FUNCTIONS
- 3 (2 HOURS LECTURE): LOCAL SEARCH AND OPTIMISATION PROBLEMS: HILL CLIMBER AND SIMULATED ANNEALING
- 4 (2 HOURS LABORATORY): USE OF LOCAL SEARCH TO SOLVE OPTIMISATION PROBLEMS
KNOWLEDGE AND UNDERSTANDING: KNOWLEDGE OF THE VARIOUS SEARCH STRATEGIES FOR SOLVING SEARCH AND OPTIMISATION PROBLEMS
APPLIED KNOWLEDGE AND UNDERSTANDING: KNOWING HOW TO CHOOSE THE MOST SUITABLE STRATEGY FOR SOLVING SEARCH AND OPTIMISATION PROBLEMS, WITH PARTICULAR REFERENCE TO THE DIGITAL ENERGY SECTOR.

MODULE 2 - TEACHING UNIT 3: MACHINE LEARNING
(LECTURE/PRACTICE/LABORATORY HOURS 6/0/4)
- 5 (2 HOURS LECTURE): SUPERVISED AND UNSUPERVISED MACHINE LEARNING. REGRESSION, CLASSIFICATION, PREDICTION
- 6 (2 HOURS LECTURE): LINEAR REGRESSION AND LOGISTIC REGRESSION
- 7 (2 HOURS LECTURE): DECISION TREES AND RANDOM FORESTS
- 8 (2 HOURS LABORATORY): APPLICATION OF MACHINE LEARNING TECHNIQUES TO REGRESSION, CLASSIFICATION AND PREDICTION PROBLEMS
- 9 (2 HOURS LABORATORY): APPLICATION OF MACHINE LEARNING TECHNIQUES TO REGRESSION, CLASSIFICATION AND PREDICTION PROBLEMS, WITH PARTICULAR REFERENCE TO THE DIGITAL ENERGY SECTOR.
KNOWLEDGE AND APPLIED COMPREHENSION: KNOWLEDGE OF VARIOUS LEARNING STRATEGIES FOR SOLVING REGRESSION, CLASSIFICATION AND PREDICTION PROBLEMS, WITH PARTICULAR REFERENCE TO THE DIGITAL ENERGY SECTOR.
APPLIED KNOWLEDGE AND UNDERSTANDING: KNOWING HOW TO CHOOSE THE MOST SUITABLE STRATEGY FOR SOLVING LEARNING PROBLEMS, WITH PARTICULAR REFERENCE TO THE DIGITAL ENERGY SECTOR.

MODULE 2 - TEACHING UNIT 4: EVOLUTIONARY COMPUTATION
(LECTURE/PRACTICE/LABORATORY HOURS 6/4/4)
- 10 (2 HOURS LECTURE): FUNDAMENTALS OF EVOLUTIONARY BIOLOGY: SELECTION, RECOMBINATION AND MUTATION - INTRODUCTION TO EVOLUTIONARY ALGORITHMS
- 11 (2 HOURS LECTURE): GENETIC ALGORITHMS AND DIFFERENTIAL EVOLUTION
- 12 (2 HOURS LECTURE): GENETIC PROGRAMMING
- 13 (2 HOURS LABORATORY): APPLICATION OF GENETIC ALGORITHMS AND DIFFERENTIAL EVOLUTION TO OPTIMISATION PROBLEMS
- 14 (2 HOURS LABORATORY): APPLICATION OF GENETIC PROGRAMMING TO OPTIMISATION AND LEARNING PROBLEMS
- 15 (2 HOURS PRACTICE): DISCUSSION OF PROJECT WORK.
- 16 (2 HOURS PRACTICE): DISCUSSION OF PROJECT WORK.
KNOWLEDGE AND ABILITY TO UNDERSTAND: TO ACQUIRE KNOWLEDGE OF THE FUNDAMENTAL ELEMENTS OF EVOLUTIONARY COMPUTATION
APPLIED KNOWLEDGE AND UNDERSTANDING: TO KNOW HOW TO APPLY EVOLUTIONARY COMPUTATION TO VARIOUS TYPES OF REAL-WORLD PROBLEMS, WITH PARTICULAR REFERENCE TO THE DIGITAL ENERGY SECTOR. TO KNOW HOW TO USE THE PYTHON FRAMEWORKS FOR THE REALISATION OF EVOLUTIONARY ALGORITHMS, AND TO CHOOSE THE PARAMETERS OF SUCH NETWORKS APPROPRIATELY.

MODULE 2 - TEACHING UNIT 5: NEURAL NETWORKS
(LECTURE/PRACTICE/LABORATORY HOURS 8/2/6)
- 17 (2 LECTURE HOURS): INTRODUCTION TO NEURAL NETWORKS: THE MCCULLOCH AND PITTS NEURON. THE ROSENBLATT PERCEPTRON. FEED-FORWARD ARCHITECTURES. FULLY-CONNECTED AND SPARSE-CONNECTED NETWORKS. MULTI-LAYER PERCEPTRON.
- 18 (2 HOURS LECTURE): FEEDFORWARD ARCHITECTURE, ACTIVATION FUNCTIONS AND THEIR CHARACTERISTICS, BACKPROPAGATION AND THE STOCHASTIC GRADIENT DESCENT ALGORITHM.
- 19 (2 HOURS LECTURE): EARLY STOPPING, MOMENTUM, ADAPTIVE LEARNING RATE, REGULARISATION.
- 20 (2 HOURS LECTURE): COMPETITIVE NEURAL NETWORKS. LEARNING VECTOR QUANTIZATION NETWORKS.
- 21 (2 HOURS LABORATORY): USE OF MLP NETWORKS IN CLASSIFICATION PROBLEMS.
- 22 (2 HOURS LABORATORY): USE OF MLP NETWORKS IN REGRESSION PROBLEMS.
- 23 (2 HOURS LABORATORY): USE OF LVQ NETWORKS IN CLASSIFICATION PROBLEMS.
- 24 (2 HOURS PRACTICE): DISCUSSION PROGRESS PROJECT WORK.
KNOWLEDGE AND UNDERSTANDING: THE NEURAL LEARNING PARADIGM. ARCHITECTURE AND FUNCTIONING OF MLP AND LVQ NETWORKS.
APPLIED KNOWLEDGE AND UNDERSTANDING: KNOWING HOW TO DESIGN MACHINE LEARNING SOLUTIONS BASED ON THE NEURAL PARADIGM, WITH PARTICULAR REFERENCE TO THE DIGITAL ENERGY SECTOR.
KNOWING HOW TO USE THE PYTHON FRAMEWORK FOR THE REALISATION OF NEURAL NETWORKS, AND HOW TO CHOOSE THE PARAMETERS OF SUCH NETWORKS APPROPRIATELY.

MODULE 2 TOTAL HOURS LECTURE/PRACTICE/LABORATORY 26/6/16

	Teaching Methods
	THE COURSE INCLUDES THEORETICAL LECTURES AND CLASSROOM EXERCISES ALSO WITH THE USAGE OF COMPUTERS. TEACHING METHODS - MODULE 2 TEACHING INCLUDES THEORETICAL LECTURES, CLASSROOM AND PRACTICAL LABORATORY EXERCISES.

	Verification of learning
	THE FINAL EXAM CONSISTS OF DISCUSSING A PROJECT WORK, AIMED AT EVALUATING: THE KNOWLEDGE AND UNDERSTANDING OF THE CONCEPTS PRESENTED DURING THE COURSE; THE ABILITY OF SOLVING STATISTICAL-DATA-ANALYSIS PROBLEMS BY APPLYING THE METHODS AND TOOLS ILLUSTRATED DURING THE COURSE. FURTHERMORE, THE PERSONAL JUDGEMENT, THE COMMUNICATION SKILLS AND THE LEARNING ABILITIES ARE ALSO EVALUATED. MODULE 2 - ASSESSMENT OF LEARNING ACHIEVEMENT OF THE TEACHING OBJECTIVES IS CERTIFIED BY PASSING AN EXAMINATION WITH A GRADE OF THIRTY.THE EXAMINATION INCLUDES THE DISCUSSION OF A PROJECT CARRIED OUT IN A GROUP (WITH GROUPS OF 2-3 PEOPLE) AND AN INDIVIDUAL ORAL INTERVIEW. THE DISCUSSION OF THE PROJECT IS AIMED AT DEMONSTRATING THE ABILITY TO APPLY KNOWLEDGE AND UNDERSTANDING BY CARRYING OUT A SIMPLE APPLICATION OF THE MACHINE LEARNING OR BIG DATA ANALYTICS TOOLS PRESENTED IN THE COURSE TO A PROBLEM PROPOSED BY THE LECTURER, WITH PARTICULAR REFERENCE TO THE DIGITAL ENERGY SECTOR. THE DISCUSSION OF THE PROJECT INVOLVES A PRACTICAL DEMONSTRATION OF HOW THE APPLICATION REALISED WORKS, THE PRESENTATION OF THE EVALUATION OF THE APPLICATION'S PERFORMANCE USING QUANTITATIVE INDICATORS, AND A DESCRIPTION OF THE TECHNICAL CHOICES MADE, POSSIBLY WITH THE AID OF SLIDES. THE ORAL INTERVIEW AIMS TO TEST THE LEVEL OF KNOWLEDGE AND UNDERSTANDING OF THE TOPICS COVERED IN THE COURSE, AS WELL AS THE STUDENT'S ABILITY TO EXPOUND.

Verification of learning

THE FINAL EXAM CONSISTS OF DISCUSSING A PROJECT WORK, AIMED AT EVALUATING: THE KNOWLEDGE AND UNDERSTANDING OF THE CONCEPTS PRESENTED DURING THE COURSE; THE ABILITY OF SOLVING STATISTICAL-DATA-ANALYSIS PROBLEMS BY APPLYING THE METHODS AND TOOLS ILLUSTRATED DURING THE COURSE. FURTHERMORE, THE PERSONAL JUDGEMENT, THE COMMUNICATION SKILLS AND THE LEARNING ABILITIES ARE ALSO EVALUATED.

MODULE 2 - ASSESSMENT OF LEARNING

ACHIEVEMENT OF THE TEACHING OBJECTIVES IS CERTIFIED BY PASSING AN EXAMINATION WITH A GRADE OF THIRTY.THE EXAMINATION INCLUDES THE DISCUSSION OF A PROJECT CARRIED OUT IN A GROUP (WITH GROUPS OF 2-3 PEOPLE) AND AN INDIVIDUAL ORAL INTERVIEW.
THE DISCUSSION OF THE PROJECT IS AIMED AT DEMONSTRATING THE ABILITY TO APPLY KNOWLEDGE AND UNDERSTANDING BY CARRYING OUT A SIMPLE APPLICATION OF THE MACHINE LEARNING OR BIG DATA ANALYTICS TOOLS PRESENTED IN THE COURSE TO A PROBLEM PROPOSED BY THE LECTURER, WITH PARTICULAR REFERENCE TO THE DIGITAL ENERGY SECTOR. THE DISCUSSION OF THE PROJECT INVOLVES A PRACTICAL DEMONSTRATION OF HOW THE APPLICATION REALISED WORKS, THE PRESENTATION OF THE EVALUATION OF THE APPLICATION'S PERFORMANCE USING QUANTITATIVE INDICATORS, AND A DESCRIPTION OF THE TECHNICAL CHOICES MADE, POSSIBLY WITH THE AID OF SLIDES.
THE ORAL INTERVIEW AIMS TO TEST THE LEVEL OF KNOWLEDGE AND UNDERSTANDING OF THE TOPICS COVERED IN THE COURSE, AS WELL AS THE STUDENT'S ABILITY TO EXPOUND.

	Texts
	AN INTRODUCTION TO STATISTICAL LEARNING, G. JAMES, D. WITTEN, T. HASTIE, R. TIBSHIRANI, SPRINGER, 2013. MODULE 2: -MODULO 2 - S. RUSSELL, P. NORVIG, ARTIFICIAL INTELLIGENCE: A MODERN APPROACH, VOLUME I, PEARSON, 4 ED., 2022 - A. BRABAZON, M. O'NEILL AND S. MCGARRAGHY, NATURAL COMPUTING ALGORITHMS, SPRINGER, 2015 SUPPLEMENTARY TEACHING MATERIAL WILL BE AVAILABLE ON THE UNIVERSITY E-LEARNING PLATFORM (HTTP://ELEARNING.UNISA.IT) ACCESSIBLE TO STUDENTS USING THEIR OWN UNIVERSITY CREDENTIALS.

Texts

AN INTRODUCTION TO STATISTICAL LEARNING,
G. JAMES, D. WITTEN, T. HASTIE, R. TIBSHIRANI,
SPRINGER, 2013.

MODULE 2:

-MODULO 2

- S. RUSSELL, P. NORVIG, ARTIFICIAL INTELLIGENCE: A MODERN APPROACH, VOLUME I, PEARSON, 4 ED., 2022
- A. BRABAZON, M. O'NEILL AND S. MCGARRAGHY, NATURAL COMPUTING ALGORITHMS, SPRINGER, 2015

SUPPLEMENTARY TEACHING MATERIAL WILL BE AVAILABLE ON THE UNIVERSITY E-LEARNING PLATFORM (HTTP://ELEARNING.UNISA.IT) ACCESSIBLE TO STUDENTS USING THEIR OWN UNIVERSITY CREDENTIALS.

	More Information
	THE COURSE IS HELD IN ENGLISH.

Lessons Timetable

BETA VERSION Data source ESSE3 [Ultima Sincronizzazione: 2024-11-18]

Alessia SAGGESE | ARTIFICIAL INTELLIGENCE