Threshold Neurons are a novel artificial neuron model inspired by the threshold mechanisms and the excitation-inhibition balance observed in biological neurons. This model is designed to enhance the computational efficiency of on-device Deep Neural Networks (DNNs), which is a significant challenge in mobile and edge computing. As tasks become increasingly complex and computational resources remain constrained, research has focused on compressing neural network structures and optimizing systems. However, limited attention has been given to optimizing the fundamental building blocks of neural networks: the neurons. Threshold Neurons offer greater efficiency than the traditional neuron paradigm by significantly reducing hardware implementation complexity. Extensive experiments validate the effectiveness of neural networks utilizing Threshold Neurons, achieving substantial power savings of 7.51x to 8.19x and area savings of 3.89x to 4.33x at the kernel level, with minimal loss in precision. Furthermore, FPGA-based implementations of these networks demonstrate 2.52x power savings and 1.75x speed enhancements at the system level.
Deep Neural Networks
Threshold Neurons
Not specified
Power savings, area savings, speed enhancements
On-device
No
Yes
Reduces hardware complexity, enhances efficiency
No
FPGA
Not specified
Not specified
Not specified
Not specified
Not specified
No
Not specified
Not specified
Not specified
Not specified
Substantial power savings
Not specified
Not specified
Not specified
Not specified
Efficient on-device inference
Not specified
Not specified
Not specified
Not specified
Not specified
Not specified
No
Not specified
Not specified
No
Not specified
Not specified
Not specified
Not specified
Not specified
No
Not specified
Not specified
0.00
Not specified
Not specified
01/01/1970
01/01/1970
Not specified
Not specified
Yes