Imbalanced Training

Imbalanced training data, where some classes are significantly under-represented compared to others, poses a major challenge for machine learning models, leading to biased predictions and poor generalization. Current research focuses on mitigating this imbalance through various techniques, including data augmentation, adversarial training to improve feature representation for under-represented groups, and novel loss functions that weight samples based on class frequency. These methods are applied across diverse domains, from recommendation systems and speech recognition to medical image analysis and large language model fine-tuning, improving model fairness and accuracy in applications where balanced datasets are difficult or impossible to obtain. Addressing imbalanced training is crucial for building robust and reliable AI systems across numerous scientific and practical applications.

Papers

October 1, 2024

Mitigating Training Imbalance in LLM Fine-Tuning via Selective Parameter Merging
Yiming Ju, Ziyi Ni, Xingrun Xing, Zhixiong Zeng, hanyu Zhao, Siqi Fan, Zheng Zhang
Training Data Class Imbalance Supervised Fine Tuning LLM Fine Tuning Parameter Selection Imbalanced Training Parameter Shift

September 27, 2023

Automatic Feature Fairness in Recommendation via Adversaries
Hengchang Hu, Yiming Cao, Zhankui He, Samson Tan, Min-Yen Kan
Adversarial Perturbation Effective Recommendation Adversary Agent Fairness Aware Generalizable Feature Space Bias Imbalanced Training

June 9, 2023

Learning Emotional Representations from Imbalanced Speech Data for Speech Emotion Recognition and Emotional Text-to-Speech
Shijun Wang, Jón Guðnason, Damian Borth
Speech Emotion Recognition Imbalanced Training

May 31, 2023

Building Manufacturing Deep Learning Models with Minimal and Imbalanced Training Data Using Domain Adaptation and Data Augmentation
Adrian Shuai Li, Elisa Bertino, Rih-Teng Wu, Ting-Yan Wu
Deep Learning Domain Adaptation Data Augmentation Data Scarcity Wafer Defect Novel Domain Adaptation Imbalanced Training

April 27, 2023

Generating images of rare concepts using pre-trained diffusion models
Dvir Samuel, Rami Ben-Ari, Simon Raviv, Nir Darshan, Gal Chechik
Image Generation Text to Image Diffusion Model Text to Image Model Pre Trained Diffusion Model Underrepresented Group Novel Concept Shot Classification Benchmark Imbalanced Training

April 13, 2023

Transfer Knowledge from Head to Tail: Uncertainty Calibration under Long-tailed Distribution
Jiahao Chen, Bing Su
Domain Adaptation Knowledge Transfer Human Head ImageNet Dataset Long Tailed Data Uncertainty Calibration Long Tailed Distribution Thin Tail Imbalanced Training

October 28, 2022

Anomaly Detection in Additive Manufacturing Processes using Supervised Classification with Imbalanced Sensor Data based on Generative Adversarial Network
Jihoon Chung, Bo Shen, Zhenyu, Kong
Generative Adversarial Network Anomaly Detection Balanced Dataset Additive Manufacturing Class Imbalanced Data Supervised Classification Imbalanced Training

October 11, 2022

Efficient Gaussian Process Model on Class-Imbalanced Datasets for Generalized Zero-Shot Learning
Changkun Ye, Nick Barnes, Lars Petersson, Russell Tsuchida
Gaussian Process Class Imbalance Neural Network Model Unseen Class Imbalanced Datasets Generalized Zero Shot Learning Imbalanced Training

September 28, 2022

Class-Imbalanced Complementary-Label Learning via Weighted Loss
Meng Wei, Yong Zhou, Zhongnian Li, Xinzheng Xu
Complementary Label Weighted Loss Complementary Label Learning Imbalanced Training

July 11, 2022

PCCT: Progressive Class-Center Triplet Loss for Imbalanced Medical Image Classification
Kanghao Chen, Weixian Lei, Rong Zhang, Shen Zhao, Wei-shi Zheng, Ruixuan Wang
Medical Image Classification Triplet Loss Imbalanced Medical Image Imbalanced Training

Imbalanced Training

Papers

Mitigating Training Imbalance in LLM Fine-Tuning via Selective Parameter Merging

Automatic Feature Fairness in Recommendation via Adversaries

Learning Emotional Representations from Imbalanced Speech Data for Speech Emotion Recognition and Emotional Text-to-Speech

Building Manufacturing Deep Learning Models with Minimal and Imbalanced Training Data Using Domain Adaptation and Data Augmentation

Generating images of rare concepts using pre-trained diffusion models

Transfer Knowledge from Head to Tail: Uncertainty Calibration under Long-tailed Distribution

Anomaly Detection in Additive Manufacturing Processes using Supervised Classification with Imbalanced Sensor Data based on Generative Adversarial Network

Efficient Gaussian Process Model on Class-Imbalanced Datasets for Generalized Zero-Shot Learning

Class-Imbalanced Complementary-Label Learning via Weighted Loss

PCCT: Progressive Class-Center Triplet Loss for Imbalanced Medical Image Classification