Intrinsic Class

Intrinsic class research focuses on identifying and leveraging the fundamental, distinguishing features of data classes, particularly when dealing with limited or imbalanced datasets. Current work explores how neural networks learn and represent these intrinsic features, examining the impact of data characteristics (e.g., image type, brain signals) and employing techniques like contrastive learning, federated learning, and specialized network architectures to enhance feature extraction and model robustness. This research is crucial for improving the performance and generalizability of machine learning models across various domains, particularly in scenarios with scarce or heterogeneous data, such as brain-computer interfaces and medical image analysis.

Papers

August 15, 2024

Pre-processing and Compression: Understanding Hidden Representation Refinement Across Imaging Domains via Intrinsic Dimension
Nicholas Konz, Maciej A. Mazurowski
Linear Compression Hidden Representation Intrinsic Dimension Pre Processing Imaging Domain Intrinsic Class

December 11, 2023

Relevant Intrinsic Feature Enhancement Network for Few-Shot Semantic Segmentation
Xiaoyi Bao, Jie Qin, Siyang Sun, Yun Zheng, Xingang Wang
Semantic Segmentation Shot Semantic Segmentation Foreground Object Enhancement Network Intrinsic Feature Intrinsic Class

August 14, 2023

Aggregating Intrinsic Information to Enhance BCI Performance through Federated Learning
Rui Liu, Yuanyuan Chen, Anran Li, Yi Ding, Han Yu, Cuntai Guan
Brain Computer Interface EEG Data Motor Imagery EEG Datasets Brain Computer Interface Performance Intrinsic Class Federated Learning Platform

March 15, 2023

SymBa: Symmetric Backpropagation-Free Contrastive Learning with Forward-Forward Algorithm for Optimizing Convergence
Heung-Chang Lee, Jeonggeun Song
Neural Network Early Stage Convergence Back Propagation Forward Forward Algorithm Plausible Learning Intrinsic Class

October 17, 2022

Understanding CNN Fragility When Learning With Imbalanced Data
Damien Dablain, Kristen N. Jacobson, Colin Bellinger, Mark Roberts, Nitesh Chawla
Deep Learning Convolutional Neural Network CNN Network Imbalanced Data Latent Class Class Diversity Intrinsic Class

July 25, 2022

Equivariance and Invariance Inductive Bias for Learning from Insufficient Data
Tan Wang, Qianru Sun, Sugiri Pranata, Karlekar Jayashree, Hanwang Zhang
LeArning Abstract Inter Part Equivariance Limited Data Data Efficient Robust Model Invariant Risk Minimization Invariance Learning Intrinsic Class

Intrinsic Class

Papers

Pre-processing and Compression: Understanding Hidden Representation Refinement Across Imaging Domains via Intrinsic Dimension

Relevant Intrinsic Feature Enhancement Network for Few-Shot Semantic Segmentation

Aggregating Intrinsic Information to Enhance BCI Performance through Federated Learning

SymBa: Symmetric Backpropagation-Free Contrastive Learning with Forward-Forward Algorithm for Optimizing Convergence

Understanding CNN Fragility When Learning With Imbalanced Data

Equivariance and Invariance Inductive Bias for Learning from Insufficient Data