Glad to have been featured by TWIML on their weekly podcast on #machinelearning. I discussed our recent work on enhancing the invariant properties of #neuralnetworks connecting areas as diverse as physics, biomechanics, differential geometry, with a lot of intuition along the way.