CLRS Algorithmic Reasoning Benchmark
收藏arXiv2022-06-04 更新2024-06-21 收录
下载链接：
https://github.com/deepmind/clrs
下载链接
链接失效反馈官方服务：
资源简介：
CLRS Algorithmic Reasoning Benchmark是由DeepMind创建的数据集，包含30个经典算法的轨迹，旨在评估神经网络的算法推理能力。数据集涵盖排序、搜索、动态规划、图算法等多种算法类型，提供输入/输出对及中间轨迹信息。该数据集支持对神经网络在不同算法任务上的泛化能力和推理性能进行测试，特别关注于超出训练分布的泛化能力。
CLRS Algorithmic Reasoning Benchmark is a dataset created by DeepMind, which contains trajectories of 30 classic algorithms and aims to evaluate the algorithmic reasoning capabilities of neural networks. The dataset covers multiple algorithm types including sorting, searching, dynamic programming, graph algorithms and others, providing input/output pairs as well as intermediate trajectory information. This dataset enables testing of the generalization ability and reasoning performance of neural networks across diverse algorithmic tasks, with a particular focus on generalization beyond the training distribution.
提供机构：
DeepMind
创建时间：
2022-05-31
