Few-Shot Neural Architecture Search
Authors: Yiyang Zhao, Linnan Wang, Yuandong Tian, Rodrigo Fonseca, Tian Guo
ICML 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | To evaluate the performance of few-shot NAS in reducing the approximation error associated with supernet and improving search efficiency of search algorithms, we conducted two types of evaluations. |
| Researcher Affiliation | Collaboration | 1Worcester Polytechnic Institute 2Brown University 3Facebook AI Research. |
| Pseudocode | No | The paper does not contain any clearly labeled pseudocode or algorithm blocks. |
| Open Source Code | No | The paper does not provide an explicit statement about releasing source code or a link to a code repository for the methodology described. |
| Open Datasets | Yes | on Image Net, ... on CIFAR10, ... We use Nas Bench-201, a public architecture dataset, which provides a unified benchmark for up-to-date NAS algorithms (Dong & Yang, 2020). Nas Bench1-shot-1 supports oneshot NAS algorithms by directly leveraging Nas Bench101 (Ying et al., 2019). |
| Dataset Splits | No | The paper mentions "validation time" and "lowest validation loss" in the context of training sub-supernets but does not specify the explicit split percentages or sample counts for the validation dataset. |
| Hardware Specification | No | The paper does not provide specific details about the hardware used, such as GPU/CPU models, memory specifications, or cloud instance types. It only mentions "GPU hours" generally. |
| Software Dependencies | No | The paper does not list specific software dependencies with their version numbers, such as programming language versions or library versions (e.g., Python 3.x, PyTorch 1.x). |
| Experiment Setup | No | The paper states, "We used the same training setup for few-shot NAS as Proxyless NAS and OFA" and similar phrases, deferring to other papers for specific experimental setup details (e.g., hyperparameters, training configurations) rather than explicitly stating them within this paper. |