Disentangled Representations via Synergy Minimization
This work addresses the challenge of extracting interpretable factors from complex data for scientists and machine learning practitioners, offering a novel approach to disentanglement.
The authors tackled the problem of learning disentangled representations by minimizing informational synergy, demonstrating that their method successfully disentangles characters from word images where independence-based methods fail.
Scientists often seek simplified representations of complex systems to facilitate prediction and understanding. If the factors comprising a representation allow us to make accurate predictions about our system, but obscuring any subset of the factors destroys our ability to make predictions, we say that the representation exhibits informational synergy. We argue that synergy is an undesirable feature in learned representations and that explicitly minimizing synergy can help disentangle the true factors of variation underlying data. We explore different ways of quantifying synergy, deriving new closed-form expressions in some cases, and then show how to modify learning to produce representations that are minimally synergistic. We introduce a benchmark task to disentangle separate characters from images of words. We demonstrate that Minimally Synergistic (MinSyn) representations correctly disentangle characters while methods relying on statistical independence fail.