This paper achieves O(1) complexity for multimillion-class classification by leveraging predefined vector systems in the latent space.
April 2, 2026
Original Paper
Using predefined vector systems to speed up neural network multimillion class classification
arXiv · 2604.00779
The Takeaway
It enables 11.6x faster inference for massive classification tasks without sacrificing accuracy. This is critical for industrial applications like large-scale product categorization or facial recognition where the number of classes traditionally bottlenecks latency.
From the abstract
Label prediction in neural networks (NNs) has O(n) complexity proportional to the number of classes. This holds true for classification using fully connected layers and cosine similarity with some set of class prototypes. In this paper we show that if NN latent space (LS) geometry is known and possesses specific properties, label prediction complexity can be significantly reduced. This is achieved by associating label prediction with the O(1) complexity closest cluster center search in a vector