Box Embeddings: An open-source library for representation learning using\n geometric structures

2021/09/10 by Tejas Chheda, Chheda, Tejas, Purujit Goyal +11 · 1 citation
Computer Science · Mathematics · #Computation and Language (cs.CL) #Digital Imaging for Blood Diseases #Domain Adaptation and Few-Shot Learning #FOS: Computer and information sciences #Machine Learning (cs.LG) #Machine Learning and Data Classification #Morphological variations and asymmetry #Natural Language Processing Techniques #Topic Modeling

paper · pdf · doi:10.48550/arxiv.2109.04997

openalex publication_date 2021/09/10 · openalex created_date 2022/07/25 · openalex updated_date 2026/07/28

Abstract

A major factor contributing to the success of modern representation learning\nis the ease of performing various vector operations. Recently, objects with\ngeometric structures (eg. distributions, complex or hyperbolic vectors, or\nregions such as cones, disks, or boxes) have been explored for their\nalternative inductive biases and additional representational capacities. In\nthis work, we introduce Box Embeddings, a Python library that enables\nresearchers to easily apply and extend probabilistic box embeddings.\n

Cited by

• Hierarchical Retrieval: The Geometry and a Pretrain-Finetune Recipe

Related

• Improving Local Identifiability in Probabilistic Box Embeddings
• Information-Geometric Set Embeddings (IGSE): From Sets to Probability Distributions
• Natural Alpha Embeddings
• Dissecting embedding method: learning higher-order structures from data
• A Probabilistic approach for Learning Embeddings without Supervision