Katkat
Back

Implement Streaming Clustering for Numbers

UberData Scientist · Onsite
MediumData ScientistOnsiteMachine Learning

Last updated

This question evaluates a data scientist's competency in online clustering for numeric streams, testing understanding of streaming algorithms, bounded-memory cluster summaries, online updates, and issues including initialization, convergence behavior, outlier handling, and concept drift.

You receive a continuous stream of numeric values. Choose an appropriate clustering algorithm and implement it so that each incoming number can be assigned to a cluster while using bounded memory.

Clarify and address the following:

  • The stream is potentially unbounded, so storing all historical values is not allowed.
  • You may assume the number of clusters k is given, or explain how you would choose it.
  • The algorithm should update cluster summaries online as new values arrive.
  • The implementation should expose at least two operations: add(value) to process a new number and get_clusters() to return the current cluster centers or summaries.
  • Discuss initialization, convergence behavior, outliers, concept drift, and how you would test correctness.