Benchmarking Functional Link Expansions for Audio Classification Tasks Simone Scardapane, Danilo Comminiello, Michele Scarpiniti, Raffaele Parisi and Aurelio Uncini Abstract Functional Link Artificial Neural Networks (FLANNs) have been exten- sively used for tasks of audio and speech classification, due to their combination of universal approximation capabilities and fast training. The performance of a FLANN, however, is known to be dependent on the specific functional link (FL) expansion that is used. In this paper, we provide an extensive benchmark of multiple FL expansions on several audio classification problems, including speech discrim- ination, genre classification, and artist recognition. Our experimental results show that a random-vector expansion is well suited for classification tasks, achieving the best accuracy in two out of three tasks. Keywords Functional links ⋅ Audio classification ⋅ Speech recognition 1 Introduction Music information retrieval (MIR) aims at efficiently retrieving songs of interest from a large database, based on the user’s requirements [5]. One of the most impor- tant tasks in MIR is automatic music classification (AMC), i.e. the capability of automatically assigning one or more labels of interest to a song, depending on its audio characteristics. Examples of labels are the genre, artist, or the perceived mood. S. Scardapane ( ✉ ) ⋅ D. Comminiello ⋅ M. Scarpiniti ⋅ R. Parisi ⋅ A. Uncini Department of Information Engineering, Electronics and Telecommunications (DIET), “Sapienza” University of Rome, Via Eudossiana 18, 00184 Rome, Italy e-mail: simone.scardapane@uniroma1.it D. Comminiello e-mail: danilo.comminiello@uniroma1.it M. Scarpiniti e-mail: michele.scarpiniti@uniroma1.it R. Parisi e-mail: raffaele.parisi@uniroma1.it A. Uncini e-mail: aurel@ieee.org © Springer International Publishing Switzerland 2016 S. Bassis et al. (eds.), Advances in Neural Networks, Smart Innovation, Systems and Technologies 54, DOI 10.1007/978-3-319-33747-0_13 133