Scalable Permutation-Aware Modeling for Temporal Set Prediction
Ashish Ranjan, Ayush Agarwal, Shalin Barot, Sushant Kumar
TL;DR
This work tackles temporal set prediction with scalability constraints by introducing PIETSP, a permutation-aware architecture that combines sequence-feature integration with permutation-equivariant and -invariant transformations. The model comprises an integrated element-sequence pathway, an element and a global evaluator, and a score-fusion mechanism, achieving a time complexity of $O\big(N(KD + D^2) + |E|D\big)$ and space complexity $O(D(|E| + K + D))$, thereby scaling linearly with key factors. Empirical results on TaFeng, DC, TaoBao, and TMS datasets show competitive or superior performance in top-$k$ metrics, along with marked improvements in inference speed and training efficiency compared to strong baselines. The approach enables practical deployment in industrial-scale temporal set prediction tasks where exact element-level forecasts and fast runtimes are essential.
Abstract
Temporal set prediction involves forecasting the elements that will appear in the next set, given a sequence of prior sets, each containing a variable number of elements. Existing methods often rely on intricate architectures with substantial computational overhead, which hampers their scalability. In this work, we introduce a novel and scalable framework that leverages permutation-equivariant and permutation-invariant transformations to efficiently model set dynamics. Our approach significantly reduces both training and inference time while maintaining competitive performance. Extensive experiments on multiple public benchmarks show that our method achieves results on par with or superior to state-of-the-art models across several evaluation metrics. These results underscore the effectiveness of our model in enabling efficient and scalable temporal set prediction.