Abstract: This paper compares the performance of activation function hardware under exponential function approximation techniques. The activation function is a key component of deep neural networks, ...

Abstract: In motion control, velocity is required to be estimated with less delay and less error, but the two are trade-offs because of an encoder's quantization noise, and they are difficult to make ...

Small. With fp16 (supported by Nvidia apex), our baseline could be trained with only 2GB GPU memory. Friendly. You may use the off-the-shelf options to apply many state-of-the-art tricks in one line.