Median Finding Algorithms and Performance Bounds

Key Insight

The RANDOMIZED-SELECT algorithm achieves an expected time complexity of O(n) for finding the median, assuming all array elements are distinct. Historically, a worst-case linear-time median-finding algorithm was developed, alongside an improved randomized version that enhances efficiency by partitioning elements based on a recursively selected small sample.

Determining the exact number of comparisons needed for median finding remains a complex problem. Early investigations established a lower bound of 2n comparisons required and an upper bound of 3n comparisons for this task. These bounds provide a foundational understanding of the problem's inherent computational cost.

Recent research has further refined these performance bounds. An improved upper bound for median finding is now slightly less than 2.95n comparisons. Concurrently, a refined lower bound has been established at (2 + ε)n comparisons, where ε represents a small positive constant, indicating ongoing progress in precisely quantifying the median-finding challenge.