Matryoshka representation learning

A Kusupati, G Bhatt, A Rege… - Advances in …, 2022 - proceedings.neurips.cc
Learned representations are a central component in modern ML systems, serving a
multitude of downstream tasks. When training such representations, it is often the case that …

Node feature extraction by self-supervised multi-scale neighborhood prediction

E Chien, WC Chang, CJ Hsieh, HF Yu, J Zhang… - arxiv preprint arxiv …, 2021 - arxiv.org
Learning on graphs has attracted significant attention in the learning community due to
numerous real-world applications. In particular, graph neural networks (GNNs), which take …

Fast multi-resolution transformer fine-tuning for extreme multi-label text classification

J Zhang, WC Chang, HF Yu… - Advances in Neural …, 2021 - proceedings.neurips.cc
Extreme multi-label text classification~(XMC) seeks to find relevant labels from an extreme
large label collection for a given text input. Many real-world applications can be formulated …

Extreme multi-label learning for semantic matching in product search

WC Chang, D Jiang, HF Yu, CH Teo, J Zhang… - Proceedings of the 27th …, 2021 - dl.acm.org
We consider the problem of semantic matching in product search: given a customer query,
retrieve all semantically related products from a huge catalog of size 100 million, or more …

Cascadexml: Rethinking transformers for end-to-end multi-resolution training in extreme multi-label classification

S Kharbanda, A Banerjee… - Advances in neural …, 2022 - proceedings.neurips.cc
Abstract Extreme Multi-label Text Classification (XMC) involves learning a classifier that can
assign an input with a subset of most relevant labels from millions of label choices. Recent …

The effect of metadata on scientific literature tagging: A cross-field cross-model study

Y Zhang, B **, Q Zhu, Y Meng, J Han - Proceedings of the ACM Web …, 2023 - dl.acm.org
Due to the exponential growth of scientific publications on the Web, there is a pressing need
to tag each paper with fine-grained topics so that researchers can track their interested fields …

Finger: Fast inference for graph-based approximate nearest neighbor search

P Chen, WC Chang, JY Jiang, HF Yu, I Dhillon… - Proceedings of the …, 2023 - dl.acm.org
Approximate K-Nearest Neighbor Search (AKNNS) has now become ubiquitous in modern
applications, such as a fast search procedure with two-tower deep learning models. Graph …

Linear classifier: An often-forgotten baseline for text classification

YC Lin, SA Chen, JJ Liu, CJ Lin - arxiv preprint arxiv:2306.07111, 2023 - arxiv.org
Large-scale pre-trained language models such as BERT are popular solutions for text
classification. Due to the superior performance of these advanced methods, nowadays …

Self-paced unified representation learning for hierarchical multi-label classification

Z Yuan, H Liu, H Zhou, D Zhang, X Zhang… - Proceedings of the …, 2024 - ojs.aaai.org
Hierarchical Multi-Label Classification (HMLC) is a well-established problem that aims at
assigning data instances to multiple classes stored in a hierarchical structure. Despite its …

Dual-encoders for extreme multi-label classification

N Gupta, D Khatri, AS Rawat, S Bhojanapalli… - arxiv preprint arxiv …, 2023 - arxiv.org
Dual-encoder (DE) models are widely used in retrieval tasks, most commonly studied on
open QA benchmarks that are often characterized by multi-class and limited training data. In …