FineMorphs: Affine-diffeomorphic sequences for regression
It addresses regression problems by applying shape analysis techniques, offering a novel approach but appears incremental in the context of existing transformation-based methods.
The paper introduces FineMorphs, a regression model using sequences of affine and diffeomorphic transformations to reshape data during learning, with experiments on UCI datasets showing favorable results compared to state-of-the-art methods and neural networks.
A multivariate regression model of affine and diffeomorphic transformation sequences - FineMorphs - is presented. Leveraging concepts from shape analysis, model states are optimally "reshaped" by diffeomorphisms generated by smooth vector fields during learning. Affine transformations and vector fields are optimized within an optimal control setting, and the model can naturally reduce (or increase) dimensionality and adapt to large datasets via suboptimal vector fields. An existence proof of solution and necessary conditions for optimality for the model are derived. Experimental results on real datasets from the UCI repository are presented, with favorable results in comparison with state-of-the-art in the literature and densely-connected neural networks in TensorFlow.