On the influence of stochastic roundoff errors and their bias on the convergence of the gradient descent method with low-precision floating-point computation

Re-assign community