arXiv:2509.12188v2 Announce Type: replace 
Abstract: The study of neural representations, both in biological and artificial systems, is increasingly revealing the importance of geometric and topological structures. Inspired by this, we introduce Event2Vec, a novel framework for learning representations of discrete event sequences. Our model leverages a simple, additive recurrent structure to learn composable, interpretable embeddings. We provide a theoretical analysis demonstrating that, under specific training objectives, our model's learned representations in a Euclidean space converge to an ideal additive structure. This ensures that the representation of a sequence is the vector sum of its constituent events, a property we term the linear additive hypothesis. To address the limitations of Euclidean geometry for hierarchical data, we also introduce a variant of our model in hyperbolic space, which is naturally suited to embedding tree-like structures with low distortion. We present experiments to validate our hypothesis. Quantitative evaluation on the Brown Corpus yields a Silhouette score of 0.0564, outperforming a Word2Vec baseline (0.0215), demonstrating the model's ability to capture structural dependencies without supervision.

تم تقديم إطار عمل جديد يسمى Event2Vec لتعلم تمثيلات تسلسلات الأحداث المنفصلة، مستفيدًا من هيكل متكرر بسيط وجمعي لإنشاء تمثيلات قابلة للتكوين وقابلة للتفسير. تُظهر التحليلات النظرية للنموذج أن التمثيلات التي تم تعلمها تتقارب إلى هيكل جمعي مثالي في الفضاء الإقليدي، مما يضمن أن تمثيل التسلسل هو مجموع المتجهات للأحداث المكونة له.

Se ha introducido un nuevo marco llamado Event2Vec para aprender representaciones de secuencias de eventos discretas, aprovechando una estructura recurrente aditiva simple para crear embeddings composables e interpretables. El análisis teórico del modelo muestra que sus representaciones aprendidas convergen a una estructura aditiva ideal en el espacio euclidiano, asegurando que la representación de una secuencia sea la suma vectorial de sus eventos constitutivos.

Un nouveau cadre appelé Event2Vec a été introduit pour apprendre des représentations de séquences d'événements discrets, en s'appuyant sur une structure récurrente additive simple pour créer des embeddings composables et interprétables. L'analyse théorique du modèle montre que ses représentations apprises convergent vers une structure additive idéale dans l'espace euclidien, garantissant que la représentation d'une séquence est la somme vectorielle de ses événements constitutifs.

A new framework called Event2Vec has been introduced for learning representations of discrete event sequences, leveraging a simple, additive recurrent structure to create composable and interpretable embeddings. The model's theoretical analysis shows that its learned representations converge to an ideal additive structure in Euclidean space, ensuring that the representation of a sequence is the vector sum of its constituent events.

Event2Vec: A Geometric Approach to Learning Composable Representations of Event Sequences

One More Thing in AI – Your Shortcut to AI Mastery

Event2Vec: A Geometric Approach to Learning Composable Representations of Event Sequences

Was this article worth reading? Share it

One More Thing in AI

LucidQuery AI

The Visualizer

Envelope

Https

Videotok

Ready to build your own newsroom?