Robust Multi-Class Classification Using Linearly Scored Categorical Cross-Entropy
Chien‐Hua Chen, Po-Hsiang Lin, Jer‐Guang Hsieh, Shu‐Ling Cheng, Jyh-Horng Jeng
Abstract
In this paper, we investigate the robust single-label multi-class classification problems in machine learning using the proposed linearly scored categorical cross-entropy for training data with wrong class labels. Deep neural networks are constructed and trained using different loss functions with various noise levels. CIFAR10 and CIFAR100 image datasets are used to calculate the estimated classification accuracy performances via 10-fold cross validation. From the simulation results, the proposed loss function may actually provide a promising alternative for robust multi-class classification problems.
Topics & Concepts
Categorical variableCross entropyArtificial intelligencePattern recognition (psychology)Computer scienceContextual image classificationEntropy (arrow of time)Class (philosophy)Artificial neural networkTraining setRobustness (evolution)Machine learningMathematicsImage (mathematics)PhysicsGeneQuantum mechanicsBiochemistryChemistryMachine Learning and Data ClassificationAnomaly Detection Techniques and ApplicationsFace and Expression Recognition