My understanding is that the optimization steps of neural networks (typically gradient descent + backpropagation) act a bit like Lloyd's algorithm. Neuron weigh

My understanding is that the optimization steps of neural networks (typically gradient descent + backpropagation) act a bit like Lloyd's algorithm. Neuron weights will push each other into place. Often, what matters is how they compare to each other, such that the error is minimized; reaching the theoretically-perfect weight is less important.