"G. d'Annunzio"
The student must have the basic notions of calculus and linear algebra.
The educational goal is the following:
Knowledge and understanding
The course aims at giving a basic knowledge of deep neural networks, and to introduce students to convolutional networks (widely used for image recognition) and to recurrent networks (often used for language recognition). To this end the course will also supply the necessary notions of probability.
Ability to apply knowledge and understanding
Et the end of the course students must be able to to carry out a group project on some application that will be assigned by the teacher; every group will write a report that will discuss the methods that have been employed and the results that have been obtained. The ability to apply knowledge and understanding will also be verified by an individual oral exam, in which the above report will be discussed.
It is expected that the ability to use feedforward neural networks is used in professional contexts, for various purposes, but possibly also in public administration or in research activities.
Deep neural networks, convolutional networks, recurrent networks
1. Deep neural networks: the stochastic gradient method, the backpropagation algorithm, the square error and cross-entropy cost functions, overfitting and regularization. 2.Convolutional networks 3. Recurrent networks The following probability topics will be discussed: law of large numbers for negatively correlated random variables, conditional expectation, conditional expectation as variabce minimizer. The programming languages Python and Keras will be introduced and used to write the code for examples of the various types of networks.
Nielsen, M.: Neural Networks and Deep Learning http://neuralnetworksanddeeplearning.com Goodfellow, I.; Bengio, Y.; Courville, A.: Deep learning, MIT Press (2016)
Lectures.
Knowledge and understanding will be verified both by a group project that will develop some application defined by the teacher and by an oral exam. At the end of the project students will write a report that will discuss the methods they have chosen and the outcomes of the project. The report will be submitted to the teacher at least one week before the oral exam and will be discussed in the oral exam.
The final grade will be expressed on a scale from 1 to 30, and will take into account both the group project and the oral exam.
The ability to apply knowledge and understanding will be verified by the above described final project.
Mail: c.costantini@unich.it Web page: https://www.dec.unich.it/home-caroli-costantini-cristina-146 For office hours check the web page or send an e-mail
