Adaptive von Mises-Fisher Likelihood Loss for Supervised Deep Time Series Hashing

2025-09-23

AI Summary

To address information loss inherent in converting real-valued embeddings to binary codes in time-series semantic hashing, this paper proposes a deep supervised hashing method based on hyperspherical embedding. The method maps features onto the unit hypersphere and models class-level semantic structure via the von Mises–Fisher (vMF) distribution. An adaptive vMF likelihood loss is introduced to explicitly maximize inter-class separation and minimize intra-class dispersion, thereby enhancing both discriminability and semantic fidelity of hash codes. By integrating deep neural networks, hyperspherical geometric constraints, and differentiable discrete optimization, the approach achieves significant improvements over state-of-the-art hashing methods across multiple time-series benchmark datasets—yielding simultaneous gains in retrieval accuracy and efficiency. The implementation is publicly available.

Technology Category

• Machine Learning: Time-Series/Data Streams
• Computer Vision: Representation Learning for Vision
• Planning, Routing, and Scheduling: Optimization of Spatio-temporal Systems

Application Category

• Graph Algorithms and Modeling for the Web: Graph embeddings and representation learning for Web-related graphs
• Semantics and Knowledge: Methods, algorithms and applications for the development of semantic models, knowledge graphs and other forms of structured data models with machine-interpretable semantics
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Indexing time series by creating compact binary representations is a fundamental task in time series data mining. Recently, deep learning-based hashing methods have proven effective for indexing time series based on semantic meaning rather than just raw similarity. The purpose of deep hashing is to map samples with the same semantic meaning to identical binary hash codes, enabling more efficient search and retrieval. Unlike other supervised representation learning methods, supervised deep hashing requires a discretization step to convert real-valued representations into binary codes, but this can induce significant information loss. In this paper, we propose a von Mises-Fisher (vMF) hashing loss. The proposed deep hashing model maps data to an M-dimensional hyperspherical space to effectively reduce information loss and models each data class as points following distinct vMF distributions. The designed loss aims to maximize the separation between each modeled vMF distribution to provide a better way to maximize the margin between each semantically different data sample. Experimental results show that our method outperforms existing baselines. The implementation is publicly available at https://github.com/jmpq97/vmf-hashing

Problem

• Indexing time series with compact binary representations for efficient search
• Reducing information loss when converting real-valued representations to binary codes
• Mapping semantically similar time series to identical hash codes effectively

Innovation

• Uses von Mises-Fisher likelihood loss
• Maps data to M-dimensional hyperspherical space
• Models each class as distinct vMF distribution