Learnability of high-dimensional targets by two-parameter models and gradient flow | Read Paper on Bytez