Active Sampling for Open-Set Classification without Initial Annotation
Authors: Zhao-Yang Liu, Sheng-Jun Huang4416-4423
AAAI 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Experimental results on multiple datasets validate the superiority of the proposed method with regard to different performance measures. and Experiments are performed on multiple datasets to validate the effectiveness of the proposed method on open-set classification. Results with regard to accuracy and F-1 measure show that our method achieves better performance on both the classification of known classes and detection of novel classes. |
| Researcher Affiliation | Academia | Zhao-Yang Liu, Sheng-Jun Huang College of Computer Science and Technology, Nanjing University of Aeronautics and Astronautics Collaborative Innovation Center of Novel Software Technology and Industrialization Nanjing 211106, China {zhaoyangliu, huangsj}@nuaa.edu.cn |
| Pseudocode | Yes | Algorithm 1 Active Sampling and Algorithm 2 The ASOCIA Algorithm |
| Open Source Code | No | The paper does not contain any explicit statements or links indicating that the source code for the described methodology is publicly available. |
| Open Datasets | Yes | Extended Yale B (Lee, Ho, and Kriegman 2005) has 28 classes, each of which has 576 images. Each image is resized to 48 42. Fashion-MNIST (Xiao, Rasul, and Vollgraf 2017) consists of 70000 examples from 10 classes, where each example is a 28 28 grayscale image of clothes or shoes. We sample 500 instances for each class from those two datasets. Coil20 (S.A.Nene, Nayar, and H.Murase 1996) contains 20 classes, each of which has 72 examples. |
| Dataset Splits | No | The paper describes training and testing sets, stating 'For Extended Yale B, half of the 5 known classes are randomly selected as the training set... At test stage, the other half of training examples from known classes together with all examples from unknown classes are used as the test set.' However, it does not explicitly mention a separate validation dataset split. |
| Hardware Specification | No | The paper does not provide any specific hardware details such as CPU/GPU models, memory, or cloud computing resources used for running its experiments. |
| Software Dependencies | No | The paper does not provide specific ancillary software details, such as library or solver names with version numbers, needed to replicate the experiment. |
| Experiment Setup | Yes | For Extended Yale B, half of the 5 known classes are randomly selected as the training set, from which the ASOCIA algorithm will actively select 150 examples from each class for annotation. and In previous experiments, the average budget number is 30 for each class in the affiliation matrix. Here we examine how the performance changes with the increase of the affiliation matrix size. |