No Fuss Distance Metric Learning using Proxies

Abstract

We address the problem of distance metric learning (DML), defined as learning a distance consistent with a notion of semantic similarity. Traditionally, for this problem supervision is expressed in the form of sets of points that follow an ordinal relationship -- an anchor point $x$ is similar to a set of positive points $Y$, and dissimilar to a set of negative points $Z$, and a loss defined over these distances is minimized. While the specifics of the optimization differ, in this work we collectively call this type of supervision Triplets and all methods that follow this pattern Triplet-Based methods. These methods are challenging to optimize. A main issue is the need for finding informative triplets, which is usually achieved by a variety of tricks such as increasing the batch size, hard or semi-hard triplet mining, etc. Even with these tricks, the convergence rate of such methods is slow. In this paper we propose to optimize the triplet loss on a different space of triplets, consisting of an anchor data point and similar and dissimilar proxy points which are learned as well. These proxies approximate the original data points, so that a triplet loss over the proxies is a tight upper bound of the original loss. This proxy-based loss is empirically better behaved. As a result, the proxy-loss improves on state-of-art results for three standard zero-shot learning datasets, by up to 15% points, while converging three times as fast as other triplet-based losses.

arXiv commentTo be presented in ICCV 2017

References in corpus (4)

• TensorFlow: Large-Scale Machine Learning on Heterogeneous Distributed Systems
• Improving the Robustness of Deep Neural Networks via Stability Training
• Metric Learning with Adaptive Density Discrimination
• Deep clustering: Discriminative embeddings for segmentation and separation

Cited by in corpus (29)

• COOT: Cooperative Hierarchical Transformer for Video-Text Representation Learning
• Modeling Uncertainty with Hedged Instance Embedding
• i-Mix: A Domain-Agnostic Strategy for Contrastive Representation Learning
• IMAE for Noise-Robust Learning: Mean Absolute Error Does Not Treat Examples Equally and Gradient Magnitude's Variance Matters
• Smart Mining for Deep Metric Learning
• Classification is a Strong Baseline for Deep Metric Learning
• Attention-based Ensemble for Deep Metric Learning
• Learning Spread-out Local Feature Descriptors
• Artificial Intelligence in Drug Discovery: Applications and Techniques
• Deep Metric Learning Meets Deep Clustering: An Novel Unsupervised Approach for Feature Embedding
• Low-Shot Learning with Imprinted Weights
• Generalization in Metric Learning: Should the Embedding Layer be the Embedding Layer?
• Instance Cross Entropy for Deep Metric Learning
• Human Motion Analysis with Deep Metric Learning
• Towards Visually Explaining Similarity Models
• Ranked List Loss for Deep Metric Learning
• Energy Confused Adversarial Metric Learning for Zero-Shot Image Retrieval and Clustering
• Calibrated neighborhood aware confidence measure for deep metric learning
• Impostor Networks for Fast Fine-Grained Recognition
• DocFace+: ID Document to Selfie Matching
• Deep Randomized Ensembles for Metric Learning
• Bias-Awareness for Zero-Shot Learning the Seen and Unseen
• Improving Image co-segmentation via Deep Metric Learning
• Learning a Unified Embedding for Visual Search at Pinterest
• Reconstruction Regularized Deep Metric Learning for Multi-label Image Classification
• Exploit Clues from Views: Self-Supervised and Regularized Learning for Multiview Object Recognition
• Diversity Transfer Network for Few-Shot Learning
• Diversified Mutual Learning for Deep Metric Learning
• Composing Text and Image for Image Retrieval - An Empirical Odyssey