Learngene: From Open-World to Your Learning Task
Authors: Qiu-Feng Wang, Xin Geng, Shu-Xia Lin, Shi-Yu Xia, Lei Qi, Ning Xu8557-8565
AAAI 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We demonstrate the effectiveness of our approach in an extensive empirical study and theoretical analysis. |
| Researcher Affiliation | Academia | MOE Key Laboratory of Computer Network and Information Integration, China School of Computer Science and Engineering, Southeast University, Nanjing 210096, China {qfwang, xgeng, shuxialin, shiyu xia, qilei, xning}@seu.edu.cn |
| Pseudocode | Yes | Algorithm 1: Adapt from the Collective Model |
| Open Source Code | No | The paper does not provide a direct link or explicit statement about the availability of its source code. |
| Open Datasets | Yes | Datasets 1) CIFAR100. This dataset consists of 60, 000 images of 100 generic object classes (Krizhevsky 2009). 2) Image Net100. This dataset has 14,197,122 images of 21841 labels (Russakovsky et al. 2014). |
| Dataset Splits | Yes | The collective model uses 60 classes as base classes, 16 classes as open-world classes, and the remaining 20 classes as novel classes, to ensure no data-interaction between the two models. |
| Hardware Specification | No | The paper does not provide specific hardware details (e.g., CPU/GPU models, memory) used for running its experiments. |
| Software Dependencies | No | The paper mentions "Mind Spore" as a deep learning computing framework in acknowledgments, but does not specify a version number or other software dependencies with versions. |
| Experiment Setup | Yes | Hyperparameters We set the learning rate to 0.001 and 0.005 when training the collective model and individual model, respectively. We fix all batch size to 32 and sample size in estimating fisher information to 128. λout is 0.9587 for CIFAR100 and 0.9733 for Imagenet100 to achieve the best performance. |