Beyond Convergence: Identifiability of Machine Learning and Deep Learning Models