Fast Core Identification

Battle Count: 66

The paper mentions crypto trading as a motivation (swapping between digital coins via USD intermediary), but the actual contribution is about one-sided matching markets (school choice, housing). The methodology has no direct application to trading strategies, portfolio optimization, or market microstructure. The matching framework could theoretically apply to order matching in exchanges, but this is not explored.

The core algorithm is relatively straightforward: construct a Markov matrix from preferences, compute the leading eigenvector via randomized SVD, and identify core members by sorting eigenvector coefficients. However, correctly implementing the preference-to-Markov-matrix mapping, handling sparse preferences, ensuring the strongly connected assumption, and achieving the claimed O(n) complexity requires careful implementation. The theoretical proofs are complex but the algorithm itself is implementable with standard linear algebra libraries.

The paper provides detailed algorithm pseudocode (Algorithm 1), complexity analysis, a numerical example with 3 agents, and experimental setup (random preferences, n from 10 to 5000, 1000 repetitions). However, no code repository is provided, and the O(1) hardware acceleration claim depends on specific hardware from [SPA+20]. The theoretical proofs are self-contained but some claims (e.g., O(1) core identification) appear to conflate preprocessing with query time.