Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
This paper introduces misi, an innovative inverted index designed for approximate nearest-neighbor search in metric spaces, utilizing a vocabulary that scales with the database size. By employing a pluggable inner index to identify each object's nearest sample points and implementing an idf-weighted shared-neighbor voting mechanism, misi achieves efficient querying while maintaining a constant expected length for posting lists. The method demonstrates significant performance improvements, achieving a 3.7x speedup over traditional graph-based approaches while effectively managing memory constraints and construction costs.
misi achieves a remarkable 3.7x speedup in approximate nearest-neighbor searches while keeping memory usage under control, making it a game-changer for large-scale vector databases.
We present misi, an inverted index for approximate nearest-neighbor search over general metric spaces whose vocabulary is a random sample of the database, of size proportional to $n$. Each object is represented by its $k_b$ nearest sample points, found by a pluggable inner index over the sample; queries are answered by an idf-weighted shared-neighbor vote followed by exact verification of $C$ candidates. The construction generalizes the NAPP index from a constant number of pivots to a linear-size vocabulary, which keeps posting lists at constant expected length $\rho = k_b/\alpha$ as $n$ grows and turns the index into a combinator: any high-recall index on $\alpha n$ points yields an index on $n$ points, for any metric. A probabilistic model gives a recall guarantee -- $k_b$ logarithmic in $n$ over the overlap gap suffices, with a verification budget the index itself estimates -- and a matching limit: the vote cannot resolve overlap differences below order $1/\sqrt{k_b}$. The design's strengths are structural: construction is $n$ independent searches -- embarrassingly parallel, deterministic, $5{,}250$ s for $10^8$ vectors on 64 cores, $3.7\times$ faster than a matched-recall graph build -- it streams under an enforced 3 GiB cap, and the portable artifact serves $10^8$ vectors from NVMe within an enforced 8 GB budget, below the working floor of the SSD-graph baseline. Its cost is query-time work: saturated graph baselines answer $6$-$16\times$ faster in RAM, and the verification budget for 0.99 recall grows as $n^{0.30}$. All results carry seeds, saturation sweeps and full configurations, are generated from run manifests, and include measured negative results. The intended applications weight construction cost, determinism, memory footprint, or black-box metrics over peak throughput: frequently rebuilt corpora, batch similarity workloads, constrained-memory serving.