Joint Class-level and Instance-level Relationship Modeling for Novel Class Discovery
Authors: Jiaying Zhou, Qingchao Chen
AAAI 2025 | Venue PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Extensive experiments on CIFAR100 and three fine-grained datasets demonstrate that our method achieves significant performance improvements compared to state-of-the-art methods. ... Experiments Experimental Setup Datasets. We conduct experiments on CIFAR100 dataset and three fine-grained datasets: Stanford Cars, CUB, and FGVC-Air Craft. |
| Researcher Affiliation | Academia | 1National Institute of Health Data Science, Peking University, Beijing, China 2Institute of Medical Technology, Peking University Health Science Center, Beijing, China 3 State Key Laboratory of General Artificial Intelligence, Peking University, Beijing, China EMAIL |
| Pseudocode | No | The paper describes its methodology in natural language and mathematical equations but does not present any structured pseudocode or algorithm blocks. |
| Open Source Code | No | The paper does not contain an explicit statement about releasing source code, nor does it provide a link to a code repository for the methodology described. |
| Open Datasets | Yes | We conduct experiments on CIFAR100 dataset and three fine-grained datasets: Stanford Cars, CUB, and FGVC-Air Craft. The number of classes and samples in each split are shown in Tab.3. ... To evaluate our method on more challenging datasets, we conduct experiments on a larger dataset, Image Net, and a more unbalanced dataset, Herbarium-19k. As shown in Tab.5, our methods outperforms the SOTA methods. |
| Dataset Splits | Yes | We divide each dataset into two parts: a labeled set and an unlabeled set. The number of classes and samples in each split are shown in Tab.3. Note that we assume that the number of unknown classes is known during the experiments. Table 3: Details of dataset splits. Dataset Labeled Unlabeled Images Classes Images Classes CIFAR100-20 40.0k 80 10.0k 20 CIFAR100-50 25.0k 50 25.0k 50 Stanford Cars 4.0k 98 4.1k 98 CUB 3.0k 100 3.0k 100 FGVC-Aircraft 3.3k 50 3.3k 50 |
| Hardware Specification | No | The paper discusses the experimental setup, including datasets and evaluation metrics, but does not specify the hardware (e.g., GPU, CPU models) used for running the experiments. |
| Software Dependencies | No | Implementation Details. For a fair comparison with the existing methods, we use the same experimental setup as previous methods. Due to space limitations, more details will be included in the supplementary materials. This statement indicates that details are external to the main paper and does not provide specific software names with version numbers. |
| Experiment Setup | No | Implementation Details. For a fair comparison with the existing methods, we use the same experimental setup as previous methods. Due to space limitations, more details will be included in the supplementary materials. This indicates that specific experimental setup details like hyperparameter values (e.g., learning rate, batch size, epochs, specific values for α, β, and τ mentioned) are not present in the main text. |