Class Separation

Class separation in machine learning focuses on improving the distinctiveness of different classes within a dataset, aiming to enhance model accuracy and robustness. Current research explores various techniques, including novel attention mechanisms, prototype-based learning, and the strategic manipulation of feature spaces to maximize inter-class distances and minimize intra-class variance, often employing neural network architectures and loss functions designed to achieve this separation. These advancements are significant because improved class separation directly translates to better classification performance, particularly in challenging scenarios like imbalanced datasets, domain adaptation, and open-set recognition, impacting diverse applications from medical image analysis to autonomous driving.

Papers

August 13, 2024

Breaking Class Barriers: Efficient Dataset Distillation via Inter-Class Feature Compensator
Xin Zhang, Jiawei Du, Ping Liu, Joey Tianyi Zhou
Synthetic Dataset Dataset Distillation Class Label Class Separation Inter Class Feature

July 23, 2024

EIANet: A Novel Domain Adaptation Approach to Maximize Class Distinction with Neural Collapse Principles
Zicheng Pan, Xiaohan Yu, Yongsheng Gao
Latent Space Source Free Domain Adaptation Fine Grained Visual Frame Attention Class Separation Social Class Novel Domain Adaptation

June 23, 2024

Predefined Prototypes for Intra-Class Separation and Disentanglement
Antonio Almudévar, Théo Mariotte, Alfonso Ortega, Marie Tahon, Luis Vicente, Antonio Miguel, Eduardo Lleida
Jina Embeddings Disentanglement Capability Visual Prototype Explainable Model Latent Prototype Prototypical Learning Class Separability Class Separation

May 22, 2024

Rethinking and Accelerating Graph Condensation: A Training-Free Approach with Class Partition
Xinyi Gao, Tong Chen, Wentao Zhang, Junliang Yu, Guanhua Ye, Quoc Viet Hung Nguyen, Hongzhi Yin
Constructive Approach Large Scale Graph Graph Condensation Class Separation Siamese Graph Class Node

April 16, 2024

Uncertainty-guided Open-Set Source-Free Unsupervised Domain Adaptation with Target-private Class Segregation
Mattia Litrico, Davide Talon, Sebastiano Battiato, Alessio Del Bue, Mario Valerio Giuffrida, Pietro Morerio
Domain Adaptation Contrastive Loss Self Supervised Contrastive Learning Novel Class Discovery Open Set Domain Adaptation Class Separation

December 14, 2023

DIRECT: Deep Active Learning under Imbalance and Label Noise
Shyam Nuggehalli, Jifan Zhang, Lalit Jain, Robert Nowak
Active Learning Label Noise Deep Active Learning Group Imbalance Active Learning Algorithm Class Separation

September 17, 2023

Conditional Mutual Information Constrained Deep Learning for Classification
En-Hui Yang, Shayan Mohajer Hamidi, Linfeng Ye, Renhao Tan, Beverly Yang
Deep Neural Network Classification Code Mutual Information Cross Entropy Standard Deep Conditional Mutual Information Class Separation

July 12, 2023

Large Class Separation is not what you need for Relational Reasoning-based OOD Detection
Lorenzo Li Lu, Giulia D'Ascenzi, Francesco Cappio Borlino, Tatiana Tommasi
Object Recognition Large Pre Trained Model Distribution Accuracy Relation Information Class Separation Inter Class Feature

May 2, 2023

Hamming Similarity and Graph Laplacians for Class Partitioning and Adversarial Image Detection
Huma Jamil, Yajing Liu, Turgay Caglar, Christina M. Cole, Nathaniel Blanchard, Christopher Peterson, Michael Kirby
Adversarial Image Graph Laplacian ReLU Activation Adversarial Detection Neural Network Representation Class Separation Hamming Space

January 31, 2023

Transfer Learning and Class Decomposition for Detecting the Cognitive Decline of Alzheimer Disease
Maha M. Alwuthaynani, Zahraa S. Abdallah, Raul Santos-Rodriguez
Convolutional Neural Network Transfer Learning Alzheimer'S Disease Medical Image Classification Recognition Task Cognitive Decline Class Separation

September 23, 2022

Understanding Open-Set Recognition by Jacobian Norm and Inter-Class Separation
Jaewoo Park, Hojin Park, Eunju Jeong, Andrew Beng Jin Teoh
Open Set Recognition Class Separation Inter Class

June 28, 2022

SHELS: Exclusive Feature Sets for Novelty Detection and Continual Learning Without Class Boundaries
Meghna Gummadi, David Kent, Jorge A. Mendez, Eric Eaton
Deep Neural Network Continual LEArning Novelty Detection Class Incremental Continual Learning Level Feature Class Separation

June 17, 2022

Maximum Class Separation as Inductive Bias in One Matrix
Tejaswi Kasarla, Gertjan J. Burghouts, Max van Spengler, Elise van der Pol, Rita Cucchiara, Pascal Mettes
Inductive Bias Softmax Function Terminator Economy Optimal Separation Class Separation Matrix Learning

March 2, 2022

Exploring Smoothness and Class-Separation for Semi-supervised Medical Image Segmentation
Yicheng Wu, Zhonghua Wu, Qianyi Wu, Zongyuan Ge, Jianfei Cai
Semi Supervised Semi Supervised Medical Image Segmentation Semi Supervised Segmentation Average Smoothness Class Separation Image Smoothing

Class Separation

Papers

Breaking Class Barriers: Efficient Dataset Distillation via Inter-Class Feature Compensator

EIANet: A Novel Domain Adaptation Approach to Maximize Class Distinction with Neural Collapse Principles

Predefined Prototypes for Intra-Class Separation and Disentanglement

Rethinking and Accelerating Graph Condensation: A Training-Free Approach with Class Partition

Uncertainty-guided Open-Set Source-Free Unsupervised Domain Adaptation with Target-private Class Segregation

DIRECT: Deep Active Learning under Imbalance and Label Noise

Conditional Mutual Information Constrained Deep Learning for Classification

Large Class Separation is not what you need for Relational Reasoning-based OOD Detection

Hamming Similarity and Graph Laplacians for Class Partitioning and Adversarial Image Detection

Transfer Learning and Class Decomposition for Detecting the Cognitive Decline of Alzheimer Disease

Understanding Open-Set Recognition by Jacobian Norm and Inter-Class Separation

SHELS: Exclusive Feature Sets for Novelty Detection and Continual Learning Without Class Boundaries

Maximum Class Separation as Inductive Bias in One Matrix

Exploring Smoothness and Class-Separation for Semi-supervised Medical Image Segmentation