The midway assessment is organised in two parts, and starts with a two-hour public seminar, followed by a closed meeting.
The purpose of the midway assessment is to evaluate the progress of the PhD project at a point when it is still possible to make small or more substantial changes.
Programme:
10:15 - 10:40 Welcome, and Presentation by Candidate Riccardo Simionato
10:45 - 11:45 Comments by, and discussion with Joshua D. Reiss
11:45 - 12:00 Plenary discussion with the public
Following the public event, there will be a closed meeting between the candidate, invited opponent, and supervisors (12:00 - 12:30).
Summary:
Modeling instruments, both acoustic and electronic, and analog audio effects represent a topic of great interest in musical acoustics and digital sound synthesis. Real-world systems present nonlinearities that provide unique soundings and behaviors that are difficult to reproduce in the digital world. Due to increasing computational capability, deep learning became a promising approach for modeling nonlinear systems, especially in circuits and analog audio effects. This dissertation focuses on how deep learning architectures can be used for modeling tasks. This question will be investigated in the context of acoustic and analog modeling. In particular, we highly consider aspects such as low latency, computational efficiency, and interactivity, which are critical for musical devices and instruments. Moreover, we aim for architectures that can be universal for a specific class or category of musical devices. The investigation considers two nonlinear time-varying cases from both acoustic and analog worlds: the compressor for the analog world and the pianoforte for the acoustic one.