Boosting Few-Shot Open-Set Recognition with Multi-Relation Margin Loss

Authors: Yongjuan Che, Yuexuan An, Hui Xue

IJCAI 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experiments on public benchmarks reveal that methods with MRM loss can improve the unknown detection of AUROC by a significant margin while correctly classifying the known classes.
Researcher Affiliation Academia 1School of Computer Science and Engineering, Southeast University, Nanjing, 210096, China 2MOE Key Laboratory of Computer Network and Information Integration (Southeast University), China {yjche, yx an, hxue}@seu.edu.cn
Pseudocode No The paper describes the method using mathematical formulations and text, but it does not include a clearly labeled pseudocode or algorithm block.
Open Source Code Yes Code is available at https://github.com/Casie-che/MRM.
Open Datasets Yes We conduct experiments on three public benchmark datasets CUB-200 [Wah et al., 2011], tiered Image Net [Ren et al., 2018], and mini Image Net [Vinyals et al., 2016] to verify the effectiveness of MRM.
Dataset Splits Yes During training, We use the validation set to select the best model. ... Following [Liu et al., 2020b], we set N = 5 and K = 1, 5 during meta-training and meta-testing.
Hardware Specification No The paper mentions using ResNet12 architecture but does not specify any hardware details like GPU models, CPU types, or memory used for the experiments.
Software Dependencies No The paper mentions using ResNet12 and an SGD optimizer but does not provide specific version numbers for any software libraries, frameworks, or programming languages used.
Experiment Setup Yes The initial learning rate is set to 0.0002 for the feature extractor and 0.002 for transformers with a multistep learning rate schedule. MRM is finetuning the feature extractor over 30 epochs with 0.0005 weight decay. ... We train the network parameters θ firstly while the radius R is fixed, then after one epoch, we calculate the radius for each class based on the embeddings extracted from the network of the latest update. ... We λ = 0.1, α = 1, β = 3, m = 1 for the loss function in Eq.(1), and ν = 0.1 for hypersphere updating in Eq.(7).