Track topics on Twitter Track topics that are important to you
We analyze the effect of synchronization on distributed stochastic gradient algorithms. By exploiting an analogy with dynamical models of biological quorum sensing, where synchronization between agents is induced through communication with a common signal, we quantify how synchronization can significantly reduce the magnitude of the noise felt by the individual distributed agents and their spatial mean. This noise reduction is in turn associated with a reduction in the smoothing of the loss function imposed by the stochastic gradient approximation. Through simulations on model nonconvex objectives, we demonstrate that coupling can stabilize higher noise levels and improve convergence. We provide a convergence analysis for strongly convex functions by deriving a bound on the expected deviation of the spatial mean of the agents from the global minimizer for an algorithm based on quorum sensing, the same algorithm with momentum, and the elastic averaging SGD (EASGD) algorithm. We discuss extensions to new algorithms that allow each agent to broadcast its current measure of success and shape the collective computation accordingly. We supplement our theoretical analysis with numerical experiments on convolutional neural networks trained on the CIFAR-10 data set, where we note a surprising regularizing property of EASGD even when applied to the nondistributed case. This observation suggests alternative second-order in-time algorithms for nondistributed optimization that are competitive with momentum methods.
This article was published in the following journal.
Name: Neural computation
This article is devoted to explore the periodic event-triggered stabilization for continuous-time stochastic systems, and to develop analysis tools/methods for stochastic periodic event-triggered cont...
Event-triggered formation control of multiagent systems under an undirected communication graph is investigated using complex-valued Laplacian. Both continuous-time and discrete-time models are consid...
First-order non-convex Riemannian optimization algorithms have gained recent popularity in structured machine learning problems including principal component analysis and low-rank matrix completion. T...
Gradient-based method has been extensively used in today's multiagent reinforcement learning (MARL). In a gradient-based MARL algorithm, each agent updates its parameterized strategy in the direction ...
In this paper, to investigate the exponential synchronization of stochastic neural networks, a new periodically intermittent discrete observation control (PIDOC) is first proposed. Different from the ...
This study evaluates the effectiveness of stochastic resonance electric stimulation on neuromuscular control and proprioception in healthy and individuals with stroke.
The primary objective of the study will be to compare intraoperative post TOF repair RVOT gradient under two different anaesthetic depths. Secondary objectives will be to follow up change ...
The purpose of this study is to examine the efficacy of a specially-constructed crib mattress that delivers gentle vibrations (stochastic vibrotactile stimulation) as a complementary, non-...
Background: Patients with bilateral vestibular hypofunction (BVH) frequently presented with dysequilibrium, dizziness and oscillopsia, leading to increased risk for fall. The mainstream fo...
The aim of this research is to study the influence of stochastic modulated vibrations on the autonomic nervous system of breast cancer patients during radiation therapy
A technique used to separate particles according to their densities in a continuous density gradient. The sample is usually mixed with a solution of known gradient materials and subjected to centrifugation. Each particle sediments to the position at which the gradient density is equal to its own. The range of the density gradient is usually greater than that of the sample particles. It is used in purifying biological materials such as proteins, nucleic acids, organelles, and cell types.
Processes that incorporate some element of randomness, used particularly to refer to a time series of random variables.
A study that uses observations at multiple time points before and after an intervention (the "interruption"), in an attempt to detect whether the intervention has had an effect significantly greater than any underlying trend over time.
Separation of particles according to density by employing a gradient of varying densities. At equilibrium each particle settles in the gradient at a point equal to its density. (McGraw-Hill Dictionary of Scientific and Technical Terms, 4th ed)
A class of statistical procedures for estimating the survival function (function of time, starting with a population 100% well at a given time and providing the percentage of the population still well at later times). The survival analysis is then used for making inferences about the effects of treatments, prognostic factors, exposures, and other covariates on the function.
Biological therapy involves the use of living organisms, substances derived from living organisms, or laboratory-produced versions of such substances to treat disease. Some biological therapies for cancer use vaccines or bacteria to stimulate the body&rs...