The concept of sublinear time algorithms, which allow for the processing of only a fraction of input data, is gaining interest in the face of increasingly large data sets. While deterministic exact algorithms exist for some problems, many require randomization and provide approximate solutions. Recent advancements have shown that sublinear time algorithms can approximate values of classical optimization problems and test properties of distributions efficiently. Techniques like the Szemeredi Regularity lemma are being used to design these algorithms. Despite progress, much is still unknown about the full potential of sublinear time algorithms. Various surveys and resources are available for those interested in this emerging field.
https://people.csail.mit.edu/ronitt/sublinear.html