Abstract:Data in the real world tends to exhibit a long-tailed label distribution, which poses great challenges for the training of neural networks in visual recognition. Existing methods tackle this problem mainly from the perspective of data quantity, i.e., the number of samples in each class. To be specific, they pay more attention to tail classes, like applying larger adjustments to the logit. However, in the training process, the quantity and difficulty of data are two intertwined and equally crucial problems. For some tail classes, the features of their instances are distinct and discriminative, which can also bring satisfactory accuracy; for some head classes, al-though with sufficient samples, the high semantic similarity with other classes and lack of discriminative features will bring bad accuracy. Based on these observations, we propose Adaptive Logit Adjustment Loss (ALA Loss) to ap-ply an adaptive adjusting term to the logit. The adaptive adjusting term is composed of two complementary factors: 1) quantity factor, which pays more attention to tail classes, and 2) difficulty factor, which adaptively pays more attention to hard instances in the training process. The difficulty factor can alleviate the over-optimization on tail yet easy in-stances and under-optimization on head yet hard instances. The synergy of the two factors can not only advance the performance on tail classes even further, but also promote the ac-curacy on head classes. Unlike previous logit adjusting methods that only concerned about data quantity, ALA Loss tackles the long-tailed problem from a more comprehensive, fine-grained and adaptive perspective. Extensive experimental re-sults show that our method achieves the state-of-the-art performance on challenging recognition benchmarks, including ImageNet-LT, iNaturalist 2018, and Places-LT.

Robust Adjusted Likelihood Function for Image Analysis.

A highly robust estimator through partially likelihood function modeling and its application in computer vision

Robust Rayleigh Regression Method for SAR Image Processing in Presence of Outliers

Robust Asymmetric Bayesian Adaptive Matrix Factorization

Label Smoothing is Robustification against Model Misspecification

Robustifying likelihoods by optimistically re-weighting data

A new robust ELM method based on a Bayesian framework with heavy-tailed distribution and weighted likelihood function.

Toward Robust RALMs: Revealing the Impact of Imperfect Retrieval on Retrieval-Augmented Language Models

An Improved Mixture Robust Probabilistic Linear Discriminant Analyzer for Fault Classification.

Adaptive Logit Adjustment Loss for Long-Tailed Visual Recognition

Misspecification-robust Sequential Neural Likelihood for Simulation-based Inference

Probabilistic Robust Regression with Adaptive Weights — a Case Study on Face Recognition

Robust GLRT approaches to signal detection in the presence of spatial-temporal uncertainty

Robust-BD Estimation and Inference for Varying-Dimensional General Linear Models

Improved Point Estimation for the Rayleigh Regression Model

Regularized Linear Discriminant Analysis Using a Nonlinear Covariance Matrix Estimator

Efficient and Robust Sparse Linear Discriminant Analysis for Data Classification

Bias Correction and Modified Profile Likelihood under the Wishart Complex Distribution

Robust Second-order LiDAR Bundle Adjustment Algorithm Using Mean Squared Group Metric

Robust Feature Matching for Remote Sensing Image Registration Via Locally Linear Transforming