Spikebench

Spikebench是一个基于开放访问神经活动数据集的尖峰序列分类基准，由法国巴黎高等师范学院认知神经科学实验室神经理论组创建。该数据集包含多个学习任务，如刺激类型分类、动物行为状态预测和神经元类型识别。数据集涵盖不同动物的不同脑区，如视网膜、前额叶皮质、运动和视觉皮质，并包含多种任务类型。通过手工制作的时间序列特征工程方法，Spikebench建立了一个强大的基准，其性能与最先进的深度学习模型相当。该数据集旨在通过多样化和挑战性的任务，评估和推动神经解码模型的发展。

Spikebench is a spike sequence classification benchmark based on open-access neural activity datasets, created by the Neural Theory Group of the Cognitive Neuroscience Laboratory at École Normale Supérieure in Paris, France. This dataset includes multiple learning tasks, such as stimulus type classification, animal behavioral state prediction, and neuronal type identification. It covers diverse brain regions across different animal species, including the retina, prefrontal cortex, motor cortex, and visual cortex, and encompasses a variety of task types. Through handcrafted time-series feature engineering, Spikebench establishes a robust benchmark whose performance is comparable to that of state-of-the-art deep learning models. This benchmark aims to evaluate and advance the development of neural decoding models via diverse and challenging tasks.