"Those who can imagine anything, can create the impossible" – A. Turing
Research
My research interests include the study and understanding of internal representations learned by (deep) neural networks, with an eye of regard to neural generative models such as Variational Autoencoders and Generative Adversarial Networks. I believe that a deeper understanding of such representation can shed a light over the overall learning process, helping us to make steps towards “opening the black box” of neural networks.
I’m also very fascinated by the creative aspects of AI, in particular with respect to the modeling and generation of musical time series. Music is a very peculiar form of art, presenting many challenges that can seriously put Machine Learning models to the test. Musical data can constitute a useful benchmark for helping us to highlight interesting properties (and relative shortcomings) of neural networks. On the other hand, I also think that studying the application of such models to Music can lead to various insights on the cognitive processes that underlie creativity and creative behaviours.
Publications
Andrea Valenti, Michele Barsotti, Raffaello Brondi, Davide Bacciu and Luca Ascari. ROS-Neuro Integration of Deep Convolutional Autoencoders for EEG Signal Compression in Real-time BCIs. (2020). Accepted at the IEEE International Conference on Systems, Man, and Cybernetics (SMC 2020). arXiv:2008.13485
Andrea Valenti, Antonio Carta and Davide Bacciu. (2020). Learning a Latent Space of Style-Aware Symbolic Music Representations by Adversarial Autoencoders. Accepted for presentation at the 24th European Conference on Artificial Intelligence (ECAI2020), ArXiv: 2001.05494
Bacciu, Davide & Chessa, Stefano & Ferro, Erina & Fortunati, Luigi & Gallicchio, Claudio & La Rosa, Davide & Llorente, Miguel & Micheli, Alessio & Palumbo, Filippo & Parodi, Oberdan & Valenti, Andrea & Vozzi, Federico. (2016). Detecting Socialization Events in Ageing People: The Experience of the DOREMI Project. 132-135. 10.1109/IE.2016.28.
Research
My research interests include the study and understanding of internal representations learned by (deep) neural networks, with an eye of regard to neural generative models such as Variational Autoencoders and Generative Adversarial Networks. I believe that a deeper understanding of such representation can shed a light over the overall learning process, helping us to make steps towards “opening the black box” of neural networks.
I’m also very fascinated by the creative aspects of AI, in particular with respect to the modeling and generation of musical time series. Music is a very peculiar form of art, presenting many challenges that can seriously put Machine Learning models to the test. Musical data can constitute a useful benchmark for helping us to highlight interesting properties (and relative shortcomings) of neural networks. On the other hand, I also think that studying the application of such models to Music can lead to various insights on the cognitive processes that underlie creativity and creative behaviours.
Publications