GPU-Accelerated Clustering for Financial Instruments at Scale

Summary

Quantitative strategies need instrument groupings that capture both ordinary co-movement and dependence during market stress, while remaining stable and inexpensive to recompute. The workflow constructs rolling absolute Pearson-correlation and tail pairwise dependence matrices (TPDM), factorizes them with nonnegative SymNMF using AdaptGrow, and derives soft loadings, hard argmax labels, and structural-break diagnostics. A trace-based formulation reduces estimated peak storage from approximately 20n2 to 4n2 bytes, while row-sharded distributed execution communicates O(nk) data; the system uses eigenspectrum-based selection between full-batch and block-stochastic AdaGrad. Across 250 synthetic windows, the setup tests planted membership changes and a joint tail-stress episode: ARI is intended to identify reclassification, whereas TPDM reveals co-crash dependence that ordinary correlation can miss. Reported factorization timings include 13.0 and 12.4 seconds for 100,000 instruments across four GB200 GPUs, and approximately two and four minutes for one million instruments across 64 GPUs on 16 nodes, not end-to-end timings.

Takeaways

1. 1

   The memory-efficient SymNMF formulation reduces estimated peak storage from approximately 20n2 to 4n2 bytes plus smaller factor buffers, allowing an approximately 100,000-instrument dependence matrix to fit on one NVIDIA GB200.

2. 2

   AdaptGrow uses the eigenspectrum to choose between full-batch AdaGrad and block-sampled gradients with SVRG correction, then expands the sample toward the full gradient if progress stalls.

3. 3

   In the synthetic experiment, about 9% of instruments lie at cluster boundaries; SymNMF preserves graded loadings for them, while hard argmax and spherical k-means assign them to single groups. Near-rank-1 TPDM inputs make hard angular partitions unstable.