Guided exploration of data summaries

Data summarization is the process of producing interpretable and representative subsets an input dataset. It usually performed following a one-shot with purpose finding best summary. A useful summary contains k individually uniform sets that are collectively diverse to be representative. Uniformity addresses interpretability diversity representativity. Finding such as difficult task when data highly large. We examine applicability Exploratory Analysis (EDA) formalize Eda4Sum, problem guided exploration summaries seeks sequentially produce connected goal maximizing their cumulative utility. Eda4Sum generalizes summarization. propose solve it one two approaches: (i) Top1Sum chooses most at each step; (ii) RLSum trains policy Deep Reinforcement Learning rewards agent for new collection step. compare these approaches top-performing EDA solutions. run extensive experiments on three large datasets. Our results demonstrate superiority our summarizing very data, need provide guidance domain experts.