1. I believe this would not possible as the OpenVINO toolkit is based on convolutional neural networks (CNN).
2. Please elaborate more on what is that you are trying to do. You can start by taking a look at the supported deep learning frameworks by the toolkit, the supported networks/layers and go from there. There have been also other users who have re-trained existing models to fit their project needs. Also, you can check the CNNLayer, CNNNetReader, and CNNNetwork for additional information.