KOALA: A Kalman Optimization Algorithm with Loss Adaptivity
This addresses the challenge of training neural networks efficiently in stochastic settings, though it appears incremental as it builds on existing optimization concepts like Momentum and Adam.
The authors tackled the problem of stochastic optimization in neural network training by interpreting the loss as a noisy observation and using Kalman filtering as an optimizer, resulting in a method called KOALA that achieves performance on par with or better than state-of-the-art algorithms across various tasks.
Optimization is often cast as a deterministic problem, where the solution is found through some iterative procedure such as gradient descent. However, when training neural networks the loss function changes over (iteration) time due to the randomized selection of a subset of the samples. This randomization turns the optimization problem into a stochastic one. We propose to consider the loss as a noisy observation with respect to some reference optimum. This interpretation of the loss allows us to adopt Kalman filtering as an optimizer, as its recursive formulation is designed to estimate unknown parameters from noisy measurements. Moreover, we show that the Kalman Filter dynamical model for the evolution of the unknown parameters can be used to capture the gradient dynamics of advanced methods such as Momentum and Adam. We call this stochastic optimization method KOALA, which is short for Kalman Optimization Algorithm with Loss Adaptivity. KOALA is an easy to implement, scalable, and efficient method to train neural networks. We provide convergence analysis and show experimentally that it yields parameter estimates that are on par with or better than existing state of the art optimization algorithms across several neural network architectures and machine learning tasks, such as computer vision and language modeling.