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


Current browse context:


Change to browse by:

References & Citations

DBLP - CS Bibliography


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

Computer Science > Machine Learning

Title: An Embarrassingly Simple Baseline for eXtreme Multi-label Prediction

Abstract: The goal of eXtreme Multi-label Learning (XML) is to design and learn a model that can automatically annotate a given data point with the most relevant subset of labels from an extremely large label set. Recently, many techniques have been proposed for XML that achieve reasonable performance on benchmark datasets. Motivated by the complexities of these methods and their subsequent training requirements, in this paper we propose a simple baseline technique for this task. Precisely, we present a global feature embedding technique for XML that can easily scale to very large datasets containing millions of data points in very high-dimensional feature space, irrespective of number of samples and labels. Next we show how an ensemble of such global embeddings can be used to achieve further boost in prediction accuracies with only linear increase in training and prediction time. During testing, we assign the labels using a weighted k-nearest neighbour classifier in the embedding space. Experiments reveal that though conceptually simple, this technique achieves quite competitive results, and has training time of less than one minute using a single CPU core with 15.6 GB RAM even for large-scale datasets such as Amazon-3M.
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:1912.08140v1 [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.