Hi OpenCog, I was wondering if people interested in robust AI would take the time to read our recent paper on robust training of ANNs, and tell me their opinion/critique of the proposal. You can get the paper from https://arxiv.org/abs/2205.13502.
The proposal comes down to the conclusion that robust learners minimize the Dirichlet energy of the hypothesis. Then, we utilize the Euler-Lagrange equations to show that robust classifiers are a solution to a partial differential equation (PDE), and further find a PDE for activation functions that would guarantee robustness according to the proposal. Feel free to contact me if you are interested in collaboration, or if you know someone that would be interested in doing so. Best, Ramin -- You received this message because you are subscribed to the Google Groups "opencog" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. To view this discussion on the web visit https://groups.google.com/d/msgid/opencog/CAHmauB6mn1sR8vjv%2B%3DGc3OH_9WoBwdeHngZBbAPSHhV76uRERQ%40mail.gmail.com.
