学术种子搜索与存储库

资源, 搜索

自由的
无需下载/安装
无需注册

Torrent files repository and directory of scientific and educational resources: Academic Torrents 是一个由社区维护的分布式存储库，它通过点对点 BitTorrent 架构促进海量科学数据集和研究论文的可扩展共享。该平台提供高速、容错的内容分发基础设施，使研究人员能够绕过传统集中式托管的带宽限制和高昂成本。

它收录了大量多学科资源，包括大规模机器学习基准、基因组序列和天文记录，同时通过加密验证和全球贡献者网络中的冗余镜像来确保数据完整性。

Access medium to huge file size. Search between datasets, scientific papers, 课程 and collections.

Sponsored and used by majors universities. A 必须 for biggest datasets for:

在全球范围内分发数TB的数据集，从而可以在不增加商业云存储经济负担的情况下独立验证实验结果。
获取机器学习和人工智能的基础训练集，确保用于模型开发的大规模数据仍然可访问且经过验证。
将开放获取的研究论文整合到技术工作流程中，减少对订阅模式的依赖，加速从学术理论到工业应用的转变。

学科： 机器学习、数据科学、学术研究、计算机视觉、自然语言处理。

Kaggle Datasets、Zenodo、Hugging Face Datasets 的免费替代方案。

学术洪流与 Kaggle 数据集、Zenodo 和 Hugging Face 数据集的比较
Academic Torrents	Kaggle 数据集	泽诺布斯	Hugging Face Datasets
Hosting model: Academic Torrents is a community-maintained distributed repository on BitTorrent infrastructure, and says it makes over 419.44 TB of research data available. Institutions such as Stanford, MIT and Cornell are listed as users.	Hosting model: Kaggle presents itself as a place to evaluate models through crowdsourced benchmarks, competitions and hackathons. Datasets are one part of that platform.	Hosting model: Zenodo is a free, open digital archive built by CERN and OpenAIRE, and its infrastructure is powered by the CERN Data Centre.	Hosting model: Each dataset is a Git repository with the files for its training, evaluation and test splits.
尺寸限制： The page states no per-dataset limit and invites users to distribute public data globally for free.		尺寸限制： Each record has 50 GB by default, with an extra allowance of up to 150 GB that can be spread across uploads.	尺寸限制： There is no per-repository size limit for datasets, but one file cannot exceed 500 GB. Splitting files below 200 GB and keeping folders under 10,000 entries is advised.
Access and tooling: The page describes academic BitTorrent infrastructure for distributing datasets and lists peer and traffic maps.	Access and tooling: The Kaggle CLI, installed with pip, can list, create, update, download or delete datasets and is released under Apache 2.0.		Access and tooling: Datasets with a supported file structure get a Dataset Viewer, and search can be filtered by language, task and licence.
成本： The page says hosting is free and offers infrastructure that can accelerate hosting at no charge.		成本： Zenodo is described as free.	成本： Free accounts get 100 GB of private storage and best-effort public storage. PRO lists 1 TB private, then paid additional storage, as listed on the Hub documentation.

涵盖的主题： 分布式存储库、BitTorrent 架构、科学数据集、研究论文、数据完整性、密码验证、冗余镜像、带宽限制、机器学习基准测试、基因组序列、天文记录、多 TB 数据集、独立验证、开放获取研究、技术工作流程、基于订阅的模型、工业应用、可扩展共享 ISO 27001、ISO 9001、ISO 19115、ISO 15489 和 ISO 20252。