Towards Effective Training of Robust Spiking Recurrent Neural Networks Under General Input Noise via Provable Analysis | IEEE Conference Publication | IEEE Xplore