Multi-domain approach identifies linked hidden factors from mixed signals
Identifying Dependent Components from Multi-domain Linear Mixtures
Machine LearningArtificial Intelligence
Summary
When you have data from different sources measuring the same thing, the hidden factors they measure can be tangled up and related. Most techniques assume these hidden factors act independently, but the authors study a way to uncover linked hidden factors across different data sources. They do this by looking at how the variation changes across sources rather than assuming independence. Their method, MuDo-CoM, uses these differences to successfully find both the hidden factors and how they mix to form the observed data.
What this means in practice
- •For medical imaging analysts: Separate underlying brain activity patterns from multi-site or multi-subject fMRI data where signals are dependent and scaled differently.
- •For industrial sensor engineers: Identify correlated source signals measured by different sensors with domain-specific effects without assuming signals are independent.
Authors
Danru Xu, Lauri Parkkonen, Sara Magliacane, Aapo Hyvärinen
Abstract
We study a linear mixing model with dependent latent components, assuming multiple data domains. Most existing models assume that the components are independent or at least uncorrelated, in line with independent component analysis (ICA). Some recent work allows for dependent components, but then makes specific assumptions such as parametric forms of dependencies, multi-view settings, or interventions, or does not recover the individual components. In contrast, we consider a multi-domain setting in which domains differ through domain-specific scalings of the components, while the distribution of the underlying latent components is the same across domains. This approach can model data collected, for example, from different sensors measuring the same process, different laboratories conducting the same experiment, different experimental conditions, or different subjects that might differ in biological or physiological factors. We show that, under sufficient domain variability, latent variables and mixing functions can be identified from second-order statistics alone. We propose the Multi-Domain Covariance Matching (MuDo-CoM) algorithm that generalizes previous methods of joint diagonalization. MuDo-CoM is validated on simulated data and a real-world fMRI dataset.