Ateneo - UPO

The course presents the machine learning and deep learning techniques for the analysis of experimental data. Besides a theoretical component, in laboratory the most advanced Python libraries for ML and DL are introduced.

Bishop, C., Pattern Recognition and Machine Learning, Springer
Goodfellow et al., Deep Learning, MIT Press

Knowledge and understanding: the course aims at providing students with solid knowledge about the most important data mining methods and instruments, focusing on the analysis of complex and real-world datasets.
Applying knowledge and understanding: given a dataset, and an adequate domain knowledge, the future graduate will be able to autonomously carry on a data analysis based on data mining and deep learning techniques, considering also the preprocessing and result evaluation phases. Such a goal will be achieved also by the development of a laboratory project, autonomously carried out by students.
Making judgments: students will acquire the ability of choosing the best strategy to manage complex problems in the fields of data analysis and machine learning. In addition, students will acquire competences of evaluation of the models obtained by applying data mining or deep learning techniques.
Communication Skills: students will learn an adequate terminology about the arguments presented in the course, in order to be able to collaborate with other experts of the field.
Learning Skills: students will acquire the ability of autonomously learning new data mining or deep learning techniques, useful in specific fields.

Formal prerequisites: none.
Substantial prerequisites: basic notions of statistics, basic notions of Python programming (data types, Controlling program flow, Functions, Lists, Classes, Exception handling)

The course is taught in class and in lab. In class the fundamental notions are presented, along with examples. The presentation is completed with discussions on the usage of different models to solve specific kinds of task for different data. In lab the student is guided in using the models seen in class. Several datasets are proposed for exercises, and it is discussed how to obtain further datasets, to be used both for exercises and for the final test.

Hands-on activity requires active participation of the students, who this way develop prcatical skills (ability to implement a program). Moreover, he active involvement leads the students to ask themselves questions (and, as a consequence, often to ask questions to the instructor); this allows a regular monitoring of the way the subject is being understood.

Students with physical disabilities, Learning Disabilities or Special Education Needs can request
specific services and tools via the Staff Sviluppo e Coordinamento Carriere e Servizi alle Studentesse

e agli Studenti, consulting the University webpage: https://www.uniupo.it/en/services/services-
students-physical-or-learning-disabilities

Students with disabilities, learning disabilities or special education needs, once they have contacted
the University Staff, can refer to the tutor in charge of the course to define the examination
modalities, concerning academic aspects.

Oral examination corcerning the analysis of one or more datasets, both with machine learning and deep learning techniques. The goal of this part is to verify that the student is autonomous in the fulfillment of a task on a real dataset.
The presentation is followed by a discussion (1-3 questions) regarding the techniques presented during the course.
The goal of the discussion is to verify that the student has acquired the basic theoretical notions discussed during the course.

- Data science pipeline (KDD process)
- preprocessing
- normalization, standardization, discretization, encoding, 1-hot encoding, label encoding
- Supervised machine learning: regression, linear regression univariate, multivariate.
Ridge, lasso regularization. Model validation. MSE, MAE, R2.
Classification
Models: decision trees, ensemble learning, random forest, Boosted trees
KNN
Logistic regression.
Evaluation metrics: Accuracy, precision, recall, Fb score.
Unsupervised ML
Clustering: K-means
DBSCAN, hierarchical clustering.
SSE, silhouette
Rand index, Mutual information.
PCA.

Neural networks:
perceptron
activation functions (sigmoid, tanh e relu)
multilayer perceptron
Use: forward propagation and backpropagation with gradient descent.
Optimizations and regularizations for deep neural networks (unstable gradient problem).
Convolutional neural networks and computer vision.
Transfer learning.
Residual neural network.
Recurrent Neural Network.
Autoencoders.
Adversarial Learning.

Python: introduction to NumPy
Scikit-learn
API scikit-learn
Preprocessing (transformers)
Classification/Regression/Clustering (estimators)
Pipelines
Model validation
train_test_split, KFold, cross_val_score, cross_val_predict
Evaluation metrics
sklearn.metrics.*
Model selection:
ParameterGrid, GridSearchCV, RandomizedSearchCV
Tensorflow and Keras for deep learning

Knowledge:
Theoretical Knowledge: knowledge of the most recent data mining techniques (pattern recognition, classification, regression, non-linear methods) and a good knowledge of the modern techniques of deep learning.
Practical knowledge: knowledge of the main data mining software. Ability of carrying on and of presenting a case study.
Competences and skills:
Application of the presented techniques: applying, using specific software, the analysis techniques to vertically solve proposed case studies.
Evaluation: critically comparing different methods and models. Choose the most suitable approach for a specific problem.
Presentation: producing a technical report about the data analysis, using an appropriate terminology about the topics of the course.
Learning: being able to use documentation autonomously, in order to acquire the competences needed to solve new problems.