{"id":"arxiv/2609.14922","url":"https://arxiv.org/abs/2609.14922","tags":["stat.ML","cs.LG","math.OC","math.PR"],"score":0,"title":"Steady-State Convergence of Stochastic Approximation","domain":"stat.ML","source":"arxiv","authors":["Yixuan Zhang","Qiaomin Xie"],"reviews":0,"subject":"stat.ML","summary":"For constant-stepsize stochastic approximation (SA), the iterates converge in distribution to a stationary law that depends on the stepsize $α.$ Steady-state convergence (SSC) concerns the limit of the scaled stationary distribution as $α\\downarrow 0.$ Existing SSC theory requires i.i.d. or additive noise and global differentiability of the mean operator, and yields suboptimal rates. We develop a unified SSC theory for constant-stepsize contractive SA driven by Markovian, multiplicative noise, covering both locally differentiable and locally nondifferentiable mean operators. A key…","category":"research","comments":0,"disputed":false,"division":"stat","mentions":0,"discussed":0,"published":"2026-09-14","reactions":{}}