Metrics from this work is now part of Latte! Please take a look there for a tested and cross-framework implementation.

---

Evaluation of Latent Space Disentanglement in the Presence of Interdependent Attributes

Supplementary materials for ISMIR 2021 LBD submission:

K. N. Watcharasupat and A. Lerch, "Evaluation of Latent Space Disentanglement in the Presence of Interdependent Attributes," submitted to the Late-Breaking Demo Session of the 22nd International Society for Music Information Retrieval Conference (ISMIR), Online, 2021. [arXiv]

Data

• NSynth: https://magenta.tensorflow.org/datasets/nsynth
• AudioCommons Timbral Models: https://github.com/AudioCommons/timbral_models

Model

Training and hyperparameters

• Adam(lr=1e-4, weight_decay=1e-6)
• ReduceLROnPlateau(factor=0.75, patience=3)
• stft(n_fft=1024, hop_length=128, window=hann_window center=True, pad_mode="constant", normalized=True)

Results

How to Cite

@article{watcharasupat2021metrics,
	title        = {Evaluation of Latent Space Disentanglement in the Presence of Interdependent Attributes},
	author       = {Karn N. Watcharasupat and Alexander Lerch},
	year         = 2021,
	month        = 11,
	booktitle    = {submitted to the Extended Abstracts for the Late-Breaking Demo Session of the 22nd International Society for Music Information Retrieval Conference},
	location     = {Online},
	publisher    = {ISMIR}
}