Statistical Methods and Applications

AGOSTINO DI CIACCIO
AGOSTINO DI CIACCIO

Learning goals.
The lab consists of the application of machine learning techniques to the analysis of images and/or textual documents.
The language used is Python 3.x with the Tensorflow package for the application of Convolutional and Recurrent Neural Networks (deep learning).

Knowledge and understanding.
Acquire the basics of machine learning techniques.
Understanding how and why to choose between alternative methods, or possibly how to combine different methods.
Ability to handle large amounts of images or text with the help of appropriate open source software.

Applying knowledge and understanding.
Students develop critical skills through the application of a wide range of statistical and machine learning models.
They also develop the critical sense through the comparison between alternative solutions to the same problem obtained using different learning logics.
They learn to critically interpret the results obtained by applying the procedures to real data sets.

Making judgements.
Students develop critical skills through the application of a wide range of machine learning and statistical models.
They also develop the critical sense through the comparison between alternative solutions to the same problem obtained using different learning logics.
They learn to critically interpret the results obtained by applying the procedures to real data sets.

Communication skills.
Students, through the study and execution of practical exercises, acquire the technical-scientific language of the discipline, which must be used appropriately in both the intermediate and final written tests and in the oral tests.
Communication skills are also developed through group activities.

Learning skills.
Students who pass the exam have learned a method of analysis that allows them to tackle the analysis of the images or text documents by machine learning techniques.

To successfully attend the laboratory it is necessary to have completed a course in statistical inference and to have a good knowledge of 1 programming language (Python, R or C). The simultaneous attendance of Big Data Analytics is recommended.

The laboratory intends to experiment some machine learning techniques for image processing or Natural Language Processing. The discussion is strongly application-oriented. Those who want a more theoretical and more extensive treatment can attend the Big-data Analytics course or Data Mining and Classification (in Italian).
Course structure.
1 - Introduction to the course
2 - Python 1 (on line course, available tutorship)
3 - Python 2 (on line course, available tutorship)
4 - Home work with Python to submit on the course page
5 - Machine learning (fundamentals)
6 - Machine learning 2 (fundamentals)
7 - Scikit-Learn: exercises
8 - Home work on Scikit-learn to submit on the course page
9 - Neural Networks (introduction)
10 - Neural Networks 2
11 - Deep Learning with Keras
12 - Tensorflow & Keras: exercises
13 - Home work on Keras to submit on the course page
14 - Image processing with NN
15 - Transfer Learning and Tuning: exercises
16 - Natural Language Processing by NN
17 - Exercises
18 - Final Home work to submit on the course page

The exam will consist in the evaluation of the home-works carried out during the course. The teacher can ask to illustrate in detail (in presence) the code prepared by the student.
Failure to attend classes or not to deliver homeworks will prevent you from passing the exam.

• Teacher's notes, articles, chapters of books, links to documentation on the WEB, software code will be distributed.
• The materials will be mostly available on the online course site (Moodle).

Deep Learning with Python (2017) F. Chollet. Manning eds.
Deep Learning (2016) Goodfellow, Bengio, Courville. MIT press.

The teaching will preferably be carried out in the presence.

As in all laboratories, attendance is required, also considering the activities that require the use of advanced software.

Learning tests will be proposed during the course. At the end, the completion of an individual project will be required.