On Maximum Entropy Linear Feature Inversion
This work addresses inconsistencies in classical linear feature inversion for researchers, but it appears incremental as it builds on existing maximum entropy methods.
The authors tackled the problem of inverting dimension-reducing linear mappings by proposing a unified maximum entropy approach, which generalizes existing methods and enables solutions for new cases like data constrained to [0, 1], with potential applications in machine learning.
We revisit the classical problem of inverting dimension-reducing linear mappings using the maximum entropy (MaxEnt) criterion. In the literature, solutions are problem-dependent, inconsistent, and use different entropy measures. We propose a new unified approach that not only specializes to the existing approaches, but offers solutions to new cases, such as when data values are constrained to [0, 1], which has new applications in machine learning.