Sorry, you need to enable JavaScript to visit this website.

BRINGING THE DISCUSSION OF MINIMA SHARPNESS TO THE AUDIO DOMAIN: A FILTER-NORMALISED EVALUATION FOR ACOUSTIC SCENE CLASSIFICATION

DOI:
10.60864/rmf5-rg52
Citation Author(s):
Manuel Milling, Andreas Triantafyllopoulos, Iosif Tsangko, Simon David Noel Rampp, Björn Wolfgang Schuller
Submitted by:
Manuel Milling
Last updated:
6 June 2024 - 10:23am
Document Type:
Presentation Slides
Document Year:
2024
Event:
Presenters:
Manuel Milling
Paper Code:
AASP-L4.2409
 

The correlation between the sharpness of loss minima and generalisation in the context of deep neural networks has been subject to discussion for a long time. Whilst mostly investigated in the context of selected benchmark data sets in the area of computer vision, we explore this aspect for the acoustic scene classification task of the DCASE2020 challenge data. Our analysis is based on two-dimensional filter-normalised visualisations and a derived sharpness measure. Our exploratory analysis shows that sharper minima tend to show better generalisation than flat minima –even more so for out-of-domain data, recorded from previously unseen devices–, thus adding to the dispute about better generalisation capabilities of flat minima. We further find that, in particular, the choice of optimisers is a main driver of the sharpness of minima and we discuss resulting limitations with respect to comparability. Our code, trained model
states and loss landscape visualisations are publicly available.

up
0 users have voted: