Swarming for Faster Convergence in Stochastic Optimization
This work addresses the challenge of efficient distributed optimization for machine learning practitioners, offering a novel approach to speed up convergence in large-scale settings.
The paper tackles the problem of slow convergence in distributed stochastic optimization by introducing a swarming-inspired method that reduces synchronization overhead, showing it outperforms centralized averaging in real-time convergence speed with error bounds that improve with network size and connectivity.
We study a distributed framework for stochastic optimization which is inspired by models of collective motion found in nature (e.g., swarming) with mild communication requirements. Specifically, we analyze a scheme in which each one of $N > 1$ independent threads, implements in a distributed and unsynchronized fashion, a stochastic gradient-descent algorithm which is perturbed by a swarming potential. Assuming the overhead caused by synchronization is not negligible, we show the swarming-based approach exhibits better performance than a centralized algorithm (based upon the average of $N$ observations) in terms of (real-time) convergence speed. We also derive an error bound that is monotone decreasing in network size and connectivity. We characterize the scheme's finite-time performances for both convex and non-convex objective functions.