Algorithm Analysis

This research area investigates the impact of different programming languages on the performance and efficiency of recursive algorithm implementations. Recursive algorithms, such as the Fibonacci sequence and the Towers of Hanoi problem, pose specific computational challenges including stack usage and duplication of subproblems, leading to varying memory and runtime efficiencies depending on implementation language characteristics and runtime environments. Understanding these differences is essential to selecting suitable languages for resource-constrained or high-performance recursive computations.

Traditional algorithm analysis often focuses on counting key comparisons, treating keys as atomic units. However, real-world data, such as strings or complex keys, require modeling comparisons at a finer granularity—symbol-by-symbol comparisons—resulting in more realistic cost models. This theme encompasses the development of analytical frameworks using probabilistic models of data sources and detailed complexity measures (e.g., symbol comparisons) for sorting and searching algorithms, offering deeper insights into algorithmic efficiency in practical environments.

This research theme addresses the challenge of monitoring execution progress in complex algorithms, especially recursive or non-linear ones often found in discrete mathematics and symbolic computation. Unlike linear algorithms, where progress can be linearly estimated, recursive computations can have widely varying execution times depending on input. Developing dynamic, bound-evaluation-based progress indicators enables users and systems to estimate completion percentages accurately, thus facilitating better resource management and user interaction during long computations.

Metaheuristic algorithms are widely applied for optimization but often lack transparent interpretability of their search dynamics. This theme explores graph-based modeling frameworks such as search trajectory networks (STNs) that represent the states and transitions visited by metaheuristics as networks. Such representations enable quantitative and visual comparison of different metaheuristics’ behaviors during continuous and combinatorial optimization, improving algorithm diagnosis, understanding, and development.

Complex financial fraud schemes such as circular trading exploit loopholes in tax systems by creating fictitious transactions to evade taxes. This theme focuses on algorithmic detection and forensic analysis using directed graph representations of trading relationships and transactions. By modeling tax actors as vertices and their commercial transactions as edges, graph theoretical and data mining techniques reveal hidden cycles and suspicious patterns indicative of circular trading, providing practical tools for authorities to combat tax evasion.

Network motifs are small, recurrent induced subgraphs that characterize functional building blocks in complex networks across domains like biology and social science. Detecting motifs is computationally expensive due to exhaustive enumeration and isomorphism testing. This theme investigates algorithmic techniques that optimize motif counting by leveraging inductive subgraph extensions, precomputed isomorphism, and pruning strategies to achieve significant speedups in motif detection, enabling analysis of large directed graphs in practical timeframes.