Embedding Semantic Hierarchy in Discrete Optimal Transport for Risk Minimization

  title={Embedding Semantic Hierarchy in Discrete Optimal Transport for Risk Minimization},
  author={Yubin Ge and Site Li and Xuyang Li and Fangfang Fan and Wanqing Xie and Jane Jia You and Xiaofeng Liu},
  journal={ICASSP 2021 - 2021 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)},
  • Yubin Ge, Site Li, +4 authors Xiaofeng Liu
  • Published 30 April 2021
  • Computer Science
  • ICASSP 2021 - 2021 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)
The widely-used cross-entropy (CE) loss-based deep networks achieved significant progress w.r.t. the classification accuracy. However, the CE loss can essentially ignore the risk of misclassification which is usually measured by the distance between the prediction and label in a semantic hierarchical tree. In this paper, we propose to incorporate the risk-aware inter-class correlation in a discrete optimal transport (DOT) training framework by configuring its ground distance matrix. The ground… 
1 Citations

Figures and Tables from this paper

Recursively Conditional Gaussian for Ordinal Unsupervised Domain Adaptation
A recursively conditional Gaussian set is adapted for ordered constraint modeling, which admits a tractable joint distribution prior to the latent space and is able to control the density of content vector that violates the poset constraints by a simple “three-sigma rule”.


Classifying web documents in a hierarchy of categories: a comprehensive study
A general hierarchical text categorization framework where the hierarchy of categories is involved in all phases of automated document classification, namely feature selection, learning and classification of a new document is proposed.
Hierarchical Semantic Risk Minimization for Large-Scale Classification.
This work constructs the balanced conservative/precipitant semantic (BCPS) risk matrix across all nodes in the semantic hierarchy with user-defined weights to adjust the tradeoff between two kinds of risks, and designs an algorithm to derive the risk-minimized predictions.
Embedding Visual Hierarchy With Deep Networks for Large-Scale Visual Recognition
By learning the tree classifier, the deep network and the visual hierarchy adaptation jointly in an end-to-end manner, the LMM algorithm can achieve higher accuracy rates on hierarchical visual recognition.
Local Bayes Risk Minimization Based Stopping Strategy for Hierarchical Classification
A Local Bayes Risk Minimization (LBRM) framework, which divides the prediction process into recursively deciding to stop or to go down at each decision node by balancing these two risks in a top-down fashion, and proposes a Weighted Tree Induced Error (WTIE) to obtain the predictions of minimum risk.
Hedging your bets: Optimizing accuracy-specificity trade-offs in large scale visual recognition
This work proposes the Dual Accuracy Reward Trade-off Search (DARTS) algorithm and proves that, under practical conditions, it converges to an optimal solution.
Wasserstein Loss With Alternative Reinforcement Learning for Severity-Aware Semantic Segmentation
  • Xiaofeng Liu, Yunhong Lu, Xiongchang Liu, Song Bai, Site Li, J. You
  • Computer Science
    IEEE Transactions on Intelligent Transportation Systems
  • 2022
A Wasserstein training framework is developed to explore the inter-class correlation by defining its ground metric as misclassification severity, and an adaptively learning scheme of the ground matrix is proposed to utilize the high-fidelity CARLA simulator.
Importance-Aware Semantic Segmentation in Self-Driving with Discrete Wasserstein Training
This paper proposes to incorporate the importance-aware inter-class correlation in a Wasserstein training framework by configuring its ground distance matrix by extending the ground metric to a linear, convex or concave increasing function w.r.t. pre-defined ground distance.
Severity-Aware Semantic Segmentation With Reinforced Wasserstein Training
This work proposes to incorporate the severity-aware inter-class correlation into the Wasserstein training framework by configuring its ground distance matrix, and can adaptively learn the ground metric in a high-fidelity simulator, following a reinforcement alternative optimization scheme.
Unimodal regularized neuron stick-breaking for ordinal classification
A simple yet efficient way to rephrase the output layer of the conventional deep neural network is proposed, in order to alleviate the effects of label noise in ordinal datasets, and a unimodal label regularization strategy is proposed.
Wasserstein Loss based Deep Object Detection
The distance metric is designed by combining the CE or binary cross-entropy (BCE) with Wasserstein distance to learn the detector considering both the discrimination and the seriousness of different misclassifications.