LCIP: Loss-Controlled Inverse Projection of High-Dimensional Image Data
This addresses a fundamental problem in data visualization and manipulation for researchers and practitioners in machine learning and data science, offering a more flexible approach to inverse projection.
The paper tackles the limitation of inverse projection methods that generate only fixed surface-like structures in data space, proposing a new method that can sweep the data space under user control, demonstrated through extensive image manipulation for style transfer.
Projections (or dimensionality reduction) methods $P$ aim to map high-dimensional data to typically 2D scatterplots for visual exploration. Inverse projection methods $P^{-1}$ aim to map this 2D space to the data space to support tasks such as data augmentation, classifier analysis, and data imputation. Current $P^{-1}$ methods suffer from a fundamental limitation -- they can only generate a fixed surface-like structure in data space, which poorly covers the richness of this space. We address this by a new method that can `sweep' the data space under user control. Our method works generically for any $P$ technique and dataset, is controlled by two intuitive user-set parameters, and is simple to implement. We demonstrate it by an extensive application involving image manipulation for style transfer.