Evaluation of Latent Space Disentanglement in the Presence of Interdependent Attributes
Not on the map. Why?
Abstract (arXiv)
Controllable music generation with deep generative models has become increasingly reliant on disentanglement learning techniques. However, current disentanglement metrics, such as mutual information gap (MIG), are often inadequate and misleading when used for evaluating latent representations in the presence of interdependent semantic attributes often encountered in real-world music datasets. In this work, we propose a dependency-aware information metric as a drop-in replacement for MIG that accounts for the inherent relationship between semantic attributes.
Key facts. Cited by 1 paper in the map (data through 7 Oct 2026). Builds on 10 of its 16 references that are in the map. First posted on arXiv 11 Oct 2021.
Builds on
10 of 16 references.
References · 10
- 2013Auto-Encoding Variational Bayescited by 3,980
- 2017Variational Inference of Disentangled Latent Concepts from Unlabeled Observationscited by 240
- 2017Neural Audio Synthesis of Musical Notes with WaveNet Autoencoderscited by 222
- 2018Isolating Sources of Disentanglement in Variational Autoencoderscited by 167
- 2018Learning Deep Disentangled Embeddings with the F-Statistic Losscited by 101
- 2019Theory and Evaluation Metrics for Learning Disentangled Representationscited by 57
- 2020Music FaderNets: Controllable Music Generation Based On High-Level Features via Low-Level Feature Modellingcited by 45
- 2018MIDI-VAE: Modeling Dynamics and Instrumentation of Music with Applications to Style Transfercited by 33
- 2019Musical Composition Style Transfer via Disentangled Timbre Representationscited by 11
- 2021Is Disentanglement enough? On Latent Representations for Controllable Music Generationcited by 7
Built on by
Citing papers per year: 2021: 1.
1 paper.