论文标题
MUBINN：用于脑电信号分类的多级二进制复发性神经网络
MuBiNN: Multi-Level Binarized Recurrent Neural Network for EEG signal Classification
论文作者
论文摘要
复发性神经网络（RNN）被广泛用于诸如脑电图分类等应用中的学习序列。由于其计算和内存密集的处理模式，复杂的RNN几乎不得部署在可穿戴设备上。通常，降低精度导致更高的效率和二元RNN作为节能解决方案。但是，幼稚的二进制方法导致脑电图分类的准确性损失显着。在本文中，我们提出了一个多级二进制的LSTM，该LSTM大大降低了计算，同时确保准确性非常接近完整的精度LSTM。我们的方法减少了3位LSTM细胞操作47*的延迟，精度损失小于0.01％。
Recurrent Neural Networks (RNN) are widely used for learning sequences in applications such as EEG classification. Complex RNNs could be hardly deployed on wearable devices due to their computation and memory-intensive processing patterns. Generally, reduction in precision leads much more efficiency and binarized RNNs are introduced as energy-efficient solutions. However, naive binarization methods lead to significant accuracy loss in EEG classification. In this paper, we propose a multi-level binarized LSTM, which significantly reduces computations whereas ensuring an accuracy pretty close to the full precision LSTM. Our method reduces the delay of the 3-bit LSTM cell operation 47* with less than 0.01% accuracy loss.
