Physics Applications of Machine Learning

Course objectives

GENERAL OBJECTIVES: The course is introductory and of interdisciplinary interest. During the lessons the most important ideas and methods of modern machine learning will be discussed. Different examples of application of machine-learning algorithms in several fields of physics will be also illustrated. The main goal of the course is to provide the student with a solid background on the topics addressed and to enable him to represent the knowledge about a specific problem domain and use reasoning techniques in order to design machine learning algorithms through a correct formulation of the problem, a critical choice of the learning algorithm and an experimental analysis to evaluate the obtained results. The course includes a substantial fraction of practical lessons with the use of computer systems (slides shown in classroom) in order to allow the student to apply the general concepts to real applications. During the practical exercises some of the machine learning algorithms discussed during the lessons will be implemented and applied to real problems. The course provides useful and expendable knowledge: - industry/technology/medical related: development and application of machine learning algorithms in leading technology sectors (automatic vision systems, autonomous driving systems, industrial automation,AI systems for robotics. etc ..); in medicine (recognition and segmentation of diagnostic images, etc.); in information technology companies (big-data analysis, search engines etc.); in finance (predictive algorithms, stock markets analysis and forecasts etc.). - basic and applied research: machine learning is a tool today widely used in both theory and experimental research, for example in high energy physics (in real-time data acquisition systems (triggers), in the interpretation of experimental results in the context of different theoretical models, etc.), in statistical physics and complex systems, in astrophysics, and in condensed matter physics, etc. SPECIFIC OBJECTIVES: A - Knowledge and understanding OF 1) To know the fundamentals of Machine Learning methods OF 1) To know the fundamentals of Deep Learning methods based on Differentiable Neural Networks OF 3) To understand the language of ML and DL B - Application skills OF 4) To be able to assemble simple programs of supervised and non-supervised learnign OF 5) To be able to solve simple science problems using ML and DL techniques OF 6) To be able to assess performances of ML and DL trained models C - Autonomy of judgment OF 8) To be able to evaluate the best way of implementing a ML and DL model D - Communication skills OF 9) To know how to communicate in written reports the results the work OF 10) To know how to discuss the characteristics and functionalities of simple computational models based on ML and DL E - Ability to learn OF 11) Being able to consult the API of the most used ML and DL software libraries OF 12) Being able to design and deploy a simple project of Applied Artificial Intelligence

Channel 1
STEFANO GIAGU Lecturers' profile

Program - Frequency - Exams

Course program
The first part of the course is dedicated to the study of the fundamental concepts and methods of Machine Learning [20 hours] The second part is dedicated to the modern Deep Learning techniques based on Differentiable Neural Networks [12 hours] The course includes computational hands-on activities [28 hours] Topics: Introduction to ML Basic of python, numpy, scipy, matplotlib Statistical Learning Probability density estimation and classification and regression algorithm based on PDE Non metric algorithms: decision trees, forests, bagging & ensambling, boosting, gradient and extreme gradient boost Dimensional reduction and data representation Clustering Linear Algorithms Neural Networks Training e advanced regularisation techniques Deep Learning and Convolutional Neural Networks Models for sequences: RNN, GRU, LSTM Deep Generative Models: Autoencoders, VAE, GANs, ... Basics of Quantum Machine Learning
Prerequisites
It is fundamental to know the basics of linear algebra, calculus, and computer algorithms and programming acquired in the first two years of bachelor's degree. It is important to have a basic knowledge of probability and statistic provided by the laboratory courses in the first two yeras of the bachelor's degree It is useful to have basic knowledge on the use of python programming.
Books
Main Textbooks: C.M. Bishop: Pattern Recognition and Machine Learning, Springer I. Goodfellow, Y. Bengio, A. Courville: Deep Learning, MIT Press C.M. Bishop: Deep Learning, Springer J. VanderPlas: Python Data Science Handbook Additional references: R.O. Duda, P.E. Hart, D.G. Stork: Pattern Classification, John Wiley & Sons Inc A.R. Webb, K.D Copsey: Statistical Pattern Recognition 3rd edition, Wiley W. McKinney, Python for Data Analysis, 2nd Edition, O'Relly
Teaching mode
The course is constituted for about 50% of lectures supported by slides projections and exercises aimed at providing basic knowledge of Machine Learning and Deep Learning. The 12 hands-on computational experiences provide some practical skills needed to develop and implement Machine Learning and Deep Learning models able to solve different problems in the field of physics and scientific research in general.
Frequency
Attendance to the lectures is not mandatory but strongly recommended. Attendance to the laboratory activities is mandatory for at least 2/3 of the hands-on sessions.
Exam mode
To pass the exam, students must have successfully completed an oral exam and completed and passed an individual or group project (with up to a maximum of 3 students) assigned during the course and to be carried out at home. To be considered passed, the oral exam and the project must receive a score of at least 18/30. The final grade is determined by the weighted average of the oral exam score (weight 60%) and the individual project score (weight 40%), with an additional 5% increment. A final grade of at least 18/30 is required to pass the exam. To achieve a score of 30/30 with honors, the student must demonstrate an excellent understanding of the topics covered in the course and show proficiency in using the necessary software tools for the development and implementation of the computational models discussed during the course. In determining the final grade, the following elements are considered: Computational practical test (60%) This is an individual assessment and in the evaluation will take into account: - The correctness of the concepts presented; - Clarity and rigor in exposition. Individual or group project (up to a maximum of 3 students) (40%) The project will be structured so that it requires a maximum of two weeks of part-time work to be completed and documented (a written report of up to 10 pages is required, along with code and dataset to reproduce the results reported in the document). The project will involve reproducing and potentially improving the results reported in a scientific paper where ML and DL methods are applied to an interesting and accessible problem for the student. The evaluation will consider: The correctness of the concepts presented; Clarity and rigor in exposition; The ability to elaborate on learned concepts in the development of original projects.
Lesson mode
The course is constituted for about 50% of lectures supported by slides projections and exercises aimed at providing basic knowledge of Machine Learning and Deep Learning. The 12 hands-on computational experiences provide some practical skills needed to develop and implement Machine Learning and Deep Learning models able to solve different problems in the field of physics and scientific research in general.
ANDREA CIARDIELLO Lecturers' profile
  • Lesson code10603330
  • Academic year2025/2026
  • CourseMathematical Sciences for Artificial Intelligence
  • CurriculumSingle curriculum
  • Year3rd year
  • Semester2nd semester
  • SSDFIS/01
  • CFU6