Towards Building A Group-based Unsupervised Representation Disentanglement Framework

Yang, Tao; Ren, Xuanchi; Wang, Yuwang; Zeng, Wenjun; Zheng, Nanning

doi:10.48550/arxiv.2102.10303

Cited by 1 publication

(1 citation statement)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Analogous to Principal Component Analysis, the goal is to find an independent or uncorrelated set of orthogonal components, where each component can be uniquely identified as the variation factor. In other words, we would like to model the latent space composed of latent variables Z in such a way that each orthogonal dimension of this space correlates with a single generative aspect of the data, often referred to in the literature as disentanglement [ 10 , 11 ]. Recent works have demonstrated the ability of VAEs, including BetaVAE [ 12 ] and JointVAE [ 13 ].…”

Section: Introductionmentioning

confidence: 99%

In Search of Disentanglement in Tandem Mass Spectrometry Datasets

Abram,

McCloskey

2023

Biomolecules

View full text Add to dashboard Cite

Generative modeling and representation learning of tandem mass spectrometry data aim to learn an interpretable and instrument-agnostic digital representation of metabolites directly from MS/MS spectra. Interpretable and instrument-agnostic digital representations would facilitate comparisons of MS/MS spectra between instrument vendors and enable better and more accurate queries of large MS/MS spectra databases for metabolite identification. In this study, we apply generative modeling and representation learning using variational autoencoders to understand the extent to which tandem mass spectra can be disentangled into their factors of generation (e.g., collision energy, ionization mode, instrument type, etc.) with minimal prior knowledge of the factors. We find that variational autoencoders can disentangle tandem mass spectra data with the proper choice of hyperparameters into meaningful latent representations aligned with known factors of variation. We develop a two-step approach to facilitate the selection of models that are disentangled, which could be applied to other complex and high-dimensional data sets.

show abstract

Section: Introductionmentioning

confidence: 99%