Energy-Efficient Neural Network Inference through Golomb-Rice Compression of Activations for Edge Devices | IEEE Conference Publication | IEEE Xplore