Data Evolution by Wittgenstein's Rule Following

The paper introduces the Wittgenstein's Rule Following (WRF) framework for evolving datasets by leveraging structural descriptors to maintain family resemblance across a sequence of historical datasets. Unlike traditional synthetic data generation, WRF extrapolates descriptor trajectories to predict future datasets, allowing for variability in sample size and feature dimensions without assuming direct transformations. This approach is significant for practitioners as it enables the generation of meaningful dataset continuations in both unsupervised and supervised contexts, enhancing the adaptability of data-driven models.