Metric Learning Loss Function

Metric learning loss functions aim to learn effective distance metrics in embedding spaces, enabling accurate similarity comparisons between data points. Current research focuses on developing loss functions that improve performance in few-shot learning scenarios and address challenges like high training complexity and overfitting, often leveraging techniques from statistical physics or incorporating contextual information to enhance robustness. These advancements are significantly impacting fields like image and natural language processing, leading to improved performance in tasks such as image retrieval and few-shot classification. The ongoing exploration of different loss functions and their impact on learned features is crucial for advancing the field and optimizing model performance.

Papers

June 27, 2023

Mean Field Theory in Deep Metric Learning
Takuya Furusawa
Loss Function Mean Field Deep Metric Learning Metric Learning Loss Function

May 15, 2023

SuSana Distancia is all you need: Enforcing class separability in metric learning via two novel distance-based loss functions for few-shot image classification
Mauricio Mendez-Ruiz, Jorge Gonzalez-Zapata, Ivan Reyes-Amezcua, Daniel Flores-Araiza, Francisco Lopez-Tiro, Andres Mendez-Vazquez, Gilberto Ochoa-Ruiz
Shot Learning Metric Learning Shot Image Classification Class Separability TinyImagenet Benchmark Metric Learning Loss Function Metric Based Few Shot Spatial Distancing

November 28, 2022

October 4, 2022

Supervised Metric Learning to Rank for Retrieval via Contextual Similarity Optimization
Christopher Liao, Theodoros Tsiligkaridis, Brian Kulis
App to App Retrieval Stable Rank Supervised Metric Contextual Similarity Metric Learning Loss Function Contextual Response Retrievability Loss Function

May 5, 2022

Do Different Deep Metric Learning Losses Lead to Similar Learned Features?
Konstantin Kobs, Michael Steininger, Andrzej Dulny, Andreas Hotho
Loss Function Saliency Map Metric Learning Loss Metric Learning Loss Function

Metric Learning Loss Function

Papers

Mean Field Theory in Deep Metric Learning

SuSana Distancia is all you need: Enforcing class separability in metric learning via two novel distance-based loss functions for few-shot image classification

Revisiting Distance Metric Learning for Few-Shot Natural Language Classification

Distance Metric Learning Loss Functions in Few-Shot Scenarios of Supervised Language Models Fine-Tuning

Supervised Metric Learning to Rank for Retrieval via Contextual Similarity Optimization

Do Different Deep Metric Learning Losses Lead to Similar Learned Features?