A Machine Learning Approach to Musically Meaningful Homogeneous Style Classification

Abstract

Recent literature has demonstrated the difficulty of classifying between composers who write in extremely similar styles (homogeneous style). Additionally, machine learning studies in this field have been exclusively of technical import with little musicological interpretability or significance. We present a supervised machine learning system which addresses the difficulty of differentiating between stylistically homogeneous composers using foundational elements of music, their complexity and interaction. Our work expands on previous style classification studies by developing more complex features as well as introducing a new class of musical features which focus on local irregularities within musical scores. We demonstrate the discriminative power of the system as applied to Haydn and Mozart's string quartets. Our results yield interpretable musicological conclusions about Haydn's and Mozart's stylistic differences while distinguishing between the composers with higher accuracy than previous studies in this domain.

Consensus Style Centralizing Auto-Encoder for Weak Style Classification

“…Recently, researchers have shown great interest in style classification, such as architectural style (Goel, Juneja, and Jawahar 2012;Xu et al 2014), music style (Herlands et al 2014), photographic style (Van Gemert 2011), manga style (Chu and Chao 2014) and fashion style (Kiapour et al 2014;Bossard et al 2013;Yamaguchi, Kiapour, and Berg 2013;Chao et al 2009). For example, in vision community, fashion style, which serves as the expressions of individual's characters and aesthetics, is related to clothing parsing (Yamaguchi et al 2012), recommendation (Wang et al 2015;Czapiewski et al 2015;Liu et al 2012a) and classification (Liu et al 2012b;Song et al 2011;Shao, Li, and Fu 2013;Di et al 2013), but essentially different.…”