Independent mechanism analysis, a new concept?
This addresses a fundamental limitation in unsupervised representation learning for researchers in machine learning and causality, though it appears incremental as it builds on existing causal mechanisms.
The paper tackles the nonidentifiability problem in nonlinear blind source separation by proposing independent mechanism analysis, which incorporates causal principles to recover identifiability and shows theoretical and empirical improvements.
Independent component analysis provides a principled framework for unsupervised representation learning, with solid theory on the identifiability of the latent code that generated the data, given only observations of mixtures thereof. Unfortunately, when the mixing is nonlinear, the model is provably nonidentifiable, since statistical independence alone does not sufficiently constrain the problem. Identifiability can be recovered in settings where additional, typically observed variables are included in the generative process. We investigate an alternative path and consider instead including assumptions reflecting the principle of independent causal mechanisms exploited in the field of causality. Specifically, our approach is motivated by thinking of each source as independently influencing the mixing process. This gives rise to a framework which we term independent mechanism analysis. We provide theoretical and empirical evidence that our approach circumvents a number of nonidentifiability issues arising in nonlinear blind source separation.