AUTHOR=Fang Hongjian , Zeng Yi , Zhao Feifei 

TITLE=Brain Inspired Sequences Production by Spiking Neural Networks With Reward-Modulated STDP

JOURNAL=Frontiers in Computational Neuroscience

VOLUME=Volume 15 - 2021

YEAR=2021

URL=https://www.frontiersin.org/journals/computational-neuroscience/articles/10.3389/fncom.2021.612041

DOI=10.3389/fncom.2021.612041

ISSN=1662-5188

ABSTRACT=Understanding and producing embedded sequences according to supra-regular grammars in language has always been considered as a high-level cognitive function of human beings, named 'syntax barrier' between humans and animals.
However, recently some neurologists through a well-designed experiment paradigm showed that macaques could be trained to produce embedded sequences involved supra-regular grammars. 
Via comparing the experimental results of macaques and preschool children, they claimed that human uniqueness may only lie in the speed and learning strategy, resulting from the chunking mechanism.
Inspired by their research, we proposed a Sequence Production Spiking Neural Network SPSNN to model the same production process, followed by memory and learning mechanism of the multi-brain region cooperation.
After experimental verification, we demonstrated SPSNN could also handle embedded sequence production tasks, striding over the 'syntax barrier'.
SPSNN used Population-Coding and STDP mechanism to realize working memory, Reward-Modulated STDP mechanism for the acquisition of supra-regular grammars, respectively.
In addition, we found the chunking mechanism indeed makes a difference to improve the robustness of our model.
As far as we know, our work is the first one towards the 'syntax barrier' in SNN filed, providing the computational foundation for further study of related underlying animals' neural mechanisms in the future. 
At the same time, we believe our work also provided exploration significance for the future interpretable neural network design.