Linear-Time Selection and Binary Heaps

Lecture 11 of CS502 Fundamentals of Algorithms presents the deterministic linear-time selection algorithm (medians of medians) and introduces the binary heap data structure. The Blum-Floyd-Pratt-Rivest-Tarjan selection algorithm guarantees a worst-case O(n) runtime by dividing n elements into groups of 5, finding the median of each group, and recursively choosing the median of these medians as a balanced pivot. This guarantees that at least 30% of elements are discarded in each recursive call, establishing the recurrence T(n) <= T(n/5) + T(7n/10) + O(n), which resolves strictly to O(n). Transitioning to priority queues, the lecture introduces the binary heap, defined as an array object that visualizes as a nearly complete binary tree filled on all levels except possibly the lowest. In a max-heap, every node satisfies the max-heap property: the value of a parent node is greater than or equal to the values of its children. Parent and child relationships are mapped efficiently using array indices: parent(i) = floor(i/2), left(i) = 2i, and right(i) = 2i + 1. Detailed analysis of recurrence relations highlights the imperative of analyzing base cases and recursive branching factors accurately. Dynamic programming state transitions are contrasted with greedy heuristics to evaluate exactness versus computational overhead. Students evaluate the trade-offs between comparison-based sorting boundaries and specialized linear-time integer distribution sorting techniques. The lesson explores intractable computational problems by examining formal polynomial reductions to known NP-complete cores. Practical code illustrations demonstrate iterative and recursive implementations, memoized table lookups, and pointer-based data structures. Geometric computational techniques including sweep-line approaches are investigated to demonstrate spatial dimension reduction in algorithms. Scholars analyze matrix multiplication parenthesization and edit distance dynamic tables to internalize two-dimensional state memoization. Network flow principles and bipartite matching formulations are examined as extensions of core graph optimization methodologies. The relationship between decision problems, certificate verification, and non-deterministic polynomial time is reinforced through formal definitions. In this lecture, students examine formal computational models, asymptotic behavior, and mathematical proofs of algorithmic correctness.