ProxyNCA++: Revisiting and Revitalizing Proxy Neighborhood Component Analysis

04/02/2020
by   Eu Wern Teh, et al.
0

We consider the problem of distance metric learning (DML), where the task is to learn an effective similarity measure between images. We revisit ProxyNCA and incorporate several enhancements. We find that low temperature scaling is a performance-critical component and explain why it works. Besides, we also discover that Global Max Pooling works better in general when compared to Global Average Pooling. Additionally, our proposed fast moving proxies also addresses small gradient issue of proxies, and this component synergizes well with low temperature scaling and Global Max Pooling. Our enhanced model, called ProxyNCA++, achieves a 22.9 percentage point average improvement of Recall@1 across four different zero-shot retrieval datasets compared to the original ProxyNCA algorithm. Furthermore, we achieve state-of-the-art results on the CUB200, Cars196, Sop, and InShop datasets, achieving Recall@1 scores of 72.2, 90.1, 81.4, and 90.9, respectively.

READ FULL TEXT

page 2

page 7

research
03/02/2021

Comparison of Methods Generalizing Max- and Average-Pooling

Max- and average-pooling are the most popular pooling methods for downsa...
research
08/26/2016

Hard Negative Mining for Metric Learning Based Zero-Shot Classification

Zero-Shot learning has been shown to be an efficient strategy for domain...
research
05/01/2020

Why and when should you pool? Analyzing Pooling in Recurrent Architectures

Pooling-based recurrent neural architectures consistently outperform the...
research
06/16/2017

A Fully Trainable Network with RNN-based Pooling

Pooling is an important component in convolutional neural networks (CNNs...
research
09/03/2021

Ordinal Pooling

In the framework of convolutional neural networks, downsampling is often...
research
11/02/2021

LogAvgExp Provides a Principled and Performant Global Pooling Operator

We seek to improve the pooling operation in neural networks, by applying...
research
01/30/2023

Proxy-based Zero-Shot Entity Linking by Effective Candidate Retrieval

A recent advancement in the domain of biomedical Entity Linking is the d...

Please sign up or login with your details

Forgot password? Click here to reset