[Paper Review] Sorting distinct integer keys using in-place associative sort

A. Emre Cetin|arXiv (Cornell University)|Sep 10, 2012

Algorithms and Data Compression11 references3 citations

TL;DR

This paper proposes an in-place associative sorting algorithm for distinct integer keys that achieves optimal O(n) time complexity under the condition m = O(n log n), where m is the key range. By modeling sorting as a cognitive memory process—practicing, storing, and retrieving—the method uses constant extra space and outperforms traditional distribution sorts like bucket and counting sort in space efficiency while maintaining linear-time performance for practical ranges.

ABSTRACT

In-place associative integer sorting technique was proposed for integer lists which requires only constant amount of additional memory replacing bucket sort, distribution counting sort and address calculation sort family of algorithms. The technique was explained by the analogy with the three main stages in the formation and retrieval of memory in cognitive neuroscience which are (i) practicing, (ii) storing and (iii) retrieval. In this study, the technique is specialized with two variants one for read-only integer keys and the other for modifiable integers. Hence, a novel algorithm is obtained that does not require additional memory other than a constant amount and sorts faster than all no matter how large is the list provided that m = O (n logn) where m is the range and n is the number of keys (or integers).

Motivation & Objective

• To develop an in-place integer sorting algorithm that avoids the high memory overhead of traditional distribution sorts like bucket and counting sort.
• To model the sorting process on cognitive neuroscience principles—practicing, storing, and retrieving—enabling efficient in-memory manipulation.
• To achieve optimal time complexity O(n) for large lists when the key range m satisfies m = O(n log n), ensuring scalability.
• To provide two variants: one for read-only keys and another for modifiable integers, each tailored to different data access patterns.
• To demonstrate that the algorithm maintains linear-time performance with constant extra space, making it suitable for large-scale in-memory sorting.

Proposed method

• The algorithm models sorting as a three-phase cognitive process: practicing (mapping distinct keys to an imaginary subspace), storing (clustering nodes with relative order), and retrieving (reconstructing sorted order using encoded cues).
• It uses a monotone bijective hash function to map integers in a predefined interval to an imaginary linear subspace within the input array, enabling in-place operations.
• A tag bit marks words as 'nodes' to distinguish them from regular integers, allowing the system to track associations and count duplicates during the 'practicing' phase.
• During the 'storing' phase, nodes are systematically clustered in the list space using their relative order, with positional information encoded in node records or associated idle integers if needed.
• In the 'retrieval' phase, each node's record or associated idle integer provides the absolute or relative position, enabling the inverse hash function to reconstruct the original key and place it in sorted order.
• The algorithm handles recursion by expanding sorted intervals from the end of the array, using returned counts to determine the next expansion start point, ensuring correct placement in the final sorted array.

Experimental results

Research questions

• RQ1 Can a sorting algorithm achieve O(n) time complexity with O(1) extra space for integer keys when m = O(n log n)?
• RQ2 How can cognitive neuroscience principles of memory formation (practicing, storing, retrieving) be applied to design an efficient in-place integer sorting algorithm?
• RQ3 What is the performance trade-off between read-only and modifiable integer variants of the associative sort under varying m/n ratios?
• RQ4 How does the algorithm maintain stability or handle instability when idle integers are displaced during node placement?
• RQ5 What is the asymptotic time complexity of the algorithm under uniform distribution and worst-case key distributions?

Key findings

• The read-only variant sorts distinct integer keys in O(n + m) time in the worst case, O(m) time on average for uniformly distributed keys, and O(n) time in the best case, all with O(1) extra space.
• The modifiable variant achieves O(n + m/(w − log n)) time complexity in the worst case, O(m/(w − log n)) in the average case, and O(n) in the best case, with the same space efficiency.
• For m/n ≤ 10, the read-only variant is more efficient due to fewer bitwise operations; beyond this ratio, the modifiable variant becomes more efficient.
• The algorithm outperforms traditional distribution sorts like bucket and counting sort in space efficiency while matching or exceeding their time performance under the condition m = O(n log n).
• The cognitive analogy—practicing, storing, retrieving—enables a systematic in-place transformation of the input array into a sorted permutation using only constant additional memory.
• The algorithm is stable in relative order for nodes but unstable for equal integers due to displacement of idle integers during node placement.