AutoBSS: An Efficient Algorithm for Block Stacking Style Search
Authors: Yikang Zhang, Jian Zhang, Zhao Zhong
NeurIPS 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | On Image Net classification task, Res Net50/Mobile Net V2/Efficient Net-B0 with our searched BSS achieve 79.29%/74.5%/77.79%, which outperform the original baselines by a large margin. More importantly, experimental results on model compression, object detection and instance segmentation show the strong generalizability of the proposed Auto BSS, and further verify the unneglectable impact of BSS on neural networks. |
| Researcher Affiliation | Industry | Yikang Zhang Huawei zhangyikang7@huawei.com Jian Zhang Huawei zhangjian157@huawei.com Zhao Zhong Huawei zorro.zhongzhao@huawei.com |
| Pseudocode | No | The paper describes the methods (Candidate Set Construction, BSSC Refining, BSS Clustering, Bayesian Optimization Based Search) in text and uses mathematical equations, but it does not provide any structured pseudocode or algorithm blocks. |
| Open Source Code | No | The paper does not contain any explicit statement about making the source code available, nor does it provide a link to a code repository. |
| Open Datasets | Yes | In this section, we conduct the main experiments of BSS search on Image Net classification task [35]... Then we conduct experiments to analyze the effectiveness of BSSC Refining and BSS Clustering. Finally, we extend the experiments to model compression, detection and instance segmentation to verify the generalization of Auto BSS... We report results on COCO dataset [44]... We report results on PSACAL VOC 2012 dataset [47] for PSPNet [48] and PSANet [49]. |
| Dataset Splits | Yes | In this section, we conduct the main experiments of BSS search on Image Net classification task [35]... We train the detection and segmentation model from scratch... We report results on PSACAL VOC 2012 dataset [47] for PSPNet [48] and PSANet [49]. Our models are pre-trained on Image Net and finetuned on train_aug (10582 images) set. |
| Hardware Specification | No | The paper discusses computing cost in general terms ('limited computing cost', 'reduce time cost') but does not provide any specific details about the hardware (e.g., GPU models, CPU types, memory) used for running the experiments. |
| Software Dependencies | No | The paper states in Appendix B.1 that 'We use PyTorch [51] with torchvision [52] for implementation.' but does not provide specific version numbers for PyTorch, torchvision, or any other software dependencies. |
| Experiment Setup | Yes | We train 120 epochs for Res Net18/50 and Mobile Net V2, 350 epochs for Efficient Net-B0/B1. We set the number of clusters as 16, 160, N /10 and N for each iteration, here N denotes the size of candidate set Ω. The detailed settings of experiments are demonstrated in the Appendix B. The detailed training settings are shown in Appendix B.1. |