We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:

Download:

Current browse context:

cs.LG

Change to browse by:

References & Citations

DBLP - CS Bibliography

Bookmark

(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Computer Science > Machine Learning

Title: On-the-fly Global Embeddings Using Random Projections for Extreme Multi-label Classification

Abstract: The goal of eXtreme Multi-label Learning (XML) is to automatically annotate a given data point with the most relevant subset of labels from an extremely large vocabulary of labels (e.g., a million labels). Lately, many attempts have been made to address this problem that achieve reasonable performance on benchmark datasets. In this paper, rather than coming-up with an altogether new method, our objective is to present and validate a simple baseline for this task. Precisely, we investigate an on-the-fly global and structure preserving feature embedding technique using random projections whose learning phase is independent of training samples and label vocabulary. Further, we show how an ensemble of multiple such learners can be used to achieve further boost in prediction accuracy with only linear increase in training and prediction time. Experiments on three public XML benchmarks show that the proposed approach obtains competitive accuracy compared with many existing methods. Additionally, it also provides around 6572x speed-up ratio in terms of training time and around 14.7x reduction in model-size compared to the closest competitors on the largest publicly available dataset.
Subjects: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV); Information Retrieval (cs.IR); Machine Learning (stat.ML)
Cite as: arXiv:1912.08140 [cs.LG]
  (or arXiv:1912.08140v2 [cs.LG] for this version)

Submission history

From: Yashaswi Verma [view email]
[v1] Tue, 17 Dec 2019 17:11:17 GMT (746kb,D)
[v2] Fri, 15 Oct 2021 06:54:43 GMT (568kb,D)

Link back to: arXiv, form interface, contact.