论文标题
钢琴vae：复音音乐的结构化表示学习
PIANOTREE VAE: Structured Representation Learning for Polyphonic Music
论文作者
论文摘要
音乐表示学习的主要方法涉及深度无监督的模型家庭变性自动编码器（VAE）。但是，大多数（如果不是全部）在此问题上的可行尝试很大程度上仅限于单声音乐。通常由更丰富的模态和更复杂的音乐结构组成，尚未在音乐表示学习的背景下解决。在这项工作中，我们提出了钢琴vae，这是VAE的一种新颖的树结构扩展，旨在适合多音音乐学习。实验证明了钢琴vae的有效性通过（i）具有有意义的多音段的潜在代码； （ii） - 除了在潜在空间中学到的体面的几何形状，更令人满意的重建； （iii） - 这种模型对下游音乐的各种作用的好处。
The dominant approach for music representation learning involves the deep unsupervised model family variational autoencoder (VAE). However, most, if not all, viable attempts on this problem have largely been limited to monophonic music. Normally composed of richer modality and more complex musical structures, the polyphonic counterpart has yet to be addressed in the context of music representation learning. In this work, we propose the PianoTree VAE, a novel tree-structure extension upon VAE aiming to fit the polyphonic music learning. The experiments prove the validity of the PianoTree VAE via (i)-semantically meaningful latent code for polyphonic segments; (ii)-more satisfiable reconstruction aside of decent geometry learned in the latent space; (iii)-this model's benefits to the variety of the downstream music generation.
