<ul><li>Researchers propose a new ternary spiking neuron model to enhance the representation capacity of binary spiking neurons in deep Q-learning.</li><li>The performance of the ternary neuron model introduced earlier is found to be inferior to binary models in deep Q-learning tasks, attributed to gradient estimation bias during training.</li><li>A novel ternary spiking neuron model is suggested to address the gradient estimation bias issue and improve performance, used as the basic computing unit in a DSQN.</li><li>Evaluation of the DSQN in seven Atari games demonstrates that the proposed ternary spiking neuron helps overcome the performance degradation of ternary neurons in Q-learning tasks, offering an improved solution for autonomous decision-making.</li></ul>

Improving Performance of Spike-based Deep Q-Learning using Ternary Neurons

Discover more