《learning discriminative features from electroence》论文精读只有摘要、结果和结论部分

’
介绍了EEG 和openmiir数据集

相似约束编码器
This tuple-based training approach 元数据

This is where deep learning techniques could help.









Thus, it can be concluded that the encoder filter has successfully extracted a component from the EEG signal that contains musicallymeaningfulinfonnation.

Trying to determine which music piece somebody listened to based on the EEG is a challenging problem.
试图通过脑电图来判断一个人听的是哪首曲子是一个很有挑战性的问题。
Attempting to do this across subjects and with a small training set

由于相似性介绍了约束编码(SCE)预训练技术

a simple convolutional filter was learned that reduced the data dimensionality by factor 64 and at the same time significantly improved the signal-to-noise ratio.

经过训练的神经网络分类器足够简单，可以让原始专家解释所学习的参数，并促进关于认知过程的发现对于学习这样简单的模型，预处理是必不可少的，因为通过基本的监督训练几乎不可能获得如图3所示的结果